main.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%pip install scikit-learn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cpu\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torchvision\n",
    "import torchvision.models as models\n",
    "import torchvision.transforms as transforms\n",
    "from torchvision.utils import save_image\n",
    "from torch.utils.data import Dataset, DataLoader\n",
    "from torchview import draw_graph  # attempts to visualize computational graphs\n",
    "import matplotlib.pyplot as plt\n",
    "import ffmpeg_quality_metrics as ffqm  # for VMAF calculation\n",
    "import sklearn\n",
    "from sklearn.metrics import mean_squared_error, mean_absolute_error, r2_score\n",
    "from PIL import Image\n",
    "import matplotlib.image as mpimg\n",
    "import cv2\n",
    "import os\n",
    "import random\n",
    "import subprocess\n",
    "import json\n",
    "import numpy as np\n",
    "import re   # for regular expressions\n",
    "import csv\n",
    "from collections import defaultdict\n",
    "import pandas as pd\n",
    "import tqdm\n",
    "import time\n",
    "import psutil\n",
    "import pickle\n",
    "import scipy.stats as st \n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "\n",
    "# set the device\n",
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
    "print(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Input video data, training set.\n",
    "\n",
    "train = \"data/trailers_train\"\n",
    "random_video = random.choice(os.listdir(train))\n",
    "print(random_video)\n",
    "\n",
    "video_path = os.path.join(train, random_video)\n",
    "print(video_path)\n",
    "\n",
    "cap = cv2.VideoCapture(video_path)\n",
    "\n",
    "cap_length = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))\n",
    "print(cap_length)\n",
    "\n",
    "random_frame = random.randrange(0, cap_length)\n",
    "print(\"printing frame number:\", random_frame)\n",
    "\n",
    "cap.set(cv2.CAP_PROP_POS_FRAMES, 1000)  # frame 1000\n",
    "\n",
    "ret, frame = cap.read()\n",
    "frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)  \n",
    "\n",
    "if ret:\n",
    "    plt.imshow(frame_rgb)\n",
    "    plt.show()\n",
    "    \n",
    "cap.release()\n",
    "cv2.destroyAllWindows()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Input video data, test set. \n",
    "\n",
    "test = \"data/trailers_test\"\n",
    "random_video = random.choice(os.listdir(test))\n",
    "print(random_video)\n",
    "\n",
    "video_path = os.path.join(test, random_video)\n",
    "print(video_path)\n",
    "\n",
    "cap = cv2.VideoCapture(video_path)\n",
    "\n",
    "cap_length = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))\n",
    "print(cap_length)\n",
    "\n",
    "random_frame = random.randrange(0, cap_length)\n",
    "print(\"printing frame number:\", random_frame)\n",
    "\n",
    "cap.set(cv2.CAP_PROP_POS_FRAMES, 1000)  # frame 1000\n",
    "\n",
    "ret, frame = cap.read()\n",
    "frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)  \n",
    "\n",
    "if ret:\n",
    "    plt.imshow(frame_rgb)\n",
    "    plt.show()\n",
    "    \n",
    "cap.release()\n",
    "cv2.destroyAllWindows()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compression of videos, train set.\n",
    "compression_levels = [23, 50, 51]       \n",
    "\n",
    "for video in os.listdir(train):\n",
    "    video_path = os.path.join(train, video)\n",
    "    if not os.path.isfile(video_path):\n",
    "        continue                   # skip the current iteration if video_path is not a file\n",
    "    \n",
    "    \n",
    "    level = random.choice(compression_levels)\n",
    "    \n",
    "    \n",
    "    # Construct the compressed file path\n",
    "    compressed_filename = f\"compressed_{level}_{video}\"\n",
    "\n",
    "    compressed_path = os.path.join(\"compressed_data_ALL\", compressed_filename)\n",
    "    \n",
    "    # Compress the video\n",
    "    subprocess.run([\"ffmpeg\", \"-i\", video_path, \"-c:v\", \"libx264\", \"-crf\", str(level), compressed_path])\n",
    "\n",
    "    vmaf_command = [\n",
    "        \"ffmpeg\",\n",
    "        \"-i\", video_path,\n",
    "        \"-i\", compressed_path,\n",
    "        \"-lavfi\", f\"[0:v][1:v]libvmaf=psnr=1:ssim=1:ms_ssim=1:log_fmt=json:log_path=vmaf_log_{level}_{video}.json\",\n",
    "        \"-f\", \"null\", \"-\"\n",
    "    ]\n",
    "    subprocess.run(vmaf_command, stdout=subprocess.PIPE, stderr=subprocess.PIPE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compression of test data.\n",
    "compression_levels = [23, 50, 51] \n",
    "\n",
    "for video in os.listdir(test):\n",
    "    video_path = os.path.join(test, video)\n",
    "    if not os.path.isfile(video_path):\n",
    "        continue                    \n",
    "    \n",
    "    level = random.choice(compression_levels)\n",
    "    \n",
    "    compressed_filename = f\"compressed_{level}_{video}\"\n",
    "\n",
    "    compressed_path = os.path.join(\"compressed_TEST_videos_ALL\", compressed_filename)\n",
    "    \n",
    "    # Compress the video\n",
    "    subprocess.run([\"ffmpeg\", \"-i\", video_path, \"-c:v\", \"libx264\", \"-crf\", str(level), compressed_path])\n",
    "\n",
    "    vmaf_command = [\n",
    "        \"ffmpeg\",\n",
    "        \"-i\", video_path,\n",
    "        \"-i\", compressed_path,\n",
    "        \"-lavfi\", f\"[0:v][1:v]libvmaf=psnr=1:ssim=1:ms_ssim=1:log_fmt=json:log_path=vmaf_log_{level}_{video}.json\",\n",
    "        \"-f\", \"null\", \"-\"\n",
    "    ]\n",
    "    subprocess.run(vmaf_command, stdout=subprocess.PIPE, stderr=subprocess.PIPE)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Control of lengths to make sure they correspond.\n",
    "print(len(os.listdir(\"data/trailers_train\")))    \n",
    "print(len(os.listdir(\"compressed_data_ALL\")))\n",
    "\n",
    "print(len(os.listdir(\"data/trailers_test\")))    \n",
    "print(len(os.listdir(\"compressed_TEST_videos_ALL\")))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Display frames with different compression levels.\n",
    "compressed_train = \"compressed_data2\"\n",
    "\n",
    "video_path_0 = \"compressed_data/compressed_0_a-hidden-life.mp4\"\n",
    "video_path_23 = \"compressed_data/compressed_23_a-hidden-life.mp4\"\n",
    "video_path_50 = \"compressed_data/compressed_50_a-hidden-life.mp4\"\n",
    "video_path_75 = \"compressed_data/compressed_75_a-hidden-life.mp4\"\n",
    "    \n",
    "cap_0 = cv2.VideoCapture(video_path_0)\n",
    "cap_75 = cv2.VideoCapture(video_path_75)\n",
    "\n",
    "cap_0.set(cv2.CAP_PROP_POS_FRAMES, 1000)\n",
    "cap_75.set(cv2.CAP_PROP_POS_FRAMES, 1000)\n",
    "\n",
    "ret_0, frame_0 = cap_0.read()\n",
    "frame_rgb_0 = cv2.cvtColor(frame_0, cv2.COLOR_BGR2RGB)   \n",
    "\n",
    "if ret_0:\n",
    "    plt.imshow(frame_rgb_0)\n",
    "    plt.show()\n",
    "    \n",
    "    \n",
    "ret_75, frame_75 = cap_75.read()\n",
    "frame_rgb_75 = cv2.cvtColor(frame_75, cv2.COLOR_BGR2RGB)   \n",
    "\n",
    "if ret_75:\n",
    "    plt.imshow(frame_rgb_75)\n",
    "    plt.show()\n",
    "    \n",
    "\n",
    "im_0 = Image.fromarray(frame_0)\n",
    "im_75 = Image.fromarray(frame_75)\n",
    "im_0.save(\"im0_TEST.png\")\n",
    "im_75.save(\"im75_TEST.png\")\n",
    "\n",
    "cap_0.release()\n",
    "cap_75.release()\n",
    "cv2.destroyAllWindows()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Contol frame rate in dataset.\n",
    "\n",
    "fps_list = []\n",
    "for video in os.listdir(train):\n",
    "    video_path = str(os.path.join(train, video))\n",
    "\n",
    "    original_cap = cv2.VideoCapture(video_path)\n",
    "    fps = int(original_cap.get(cv2.CAP_PROP_FPS))\n",
    "    original_frame_count = int(original_cap.get(cv2.CAP_PROP_FRAME_COUNT))\n",
    "    fps_list.append(fps)\n",
    "    \n",
    "    print(video_path, fps)\n",
    "    original_cap.release()\n",
    "    \n",
    "# print(min(fps_list))\n",
    "# print(max(fps_list))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Dataset with compressed frames, ≈ 1 frame / 10 seconds, png format neccessary.\n",
    "\n",
    "compressed_train = \"compressed_data_ALL\"\n",
    "nr = 1\n",
    "\n",
    "video_files = sorted(os.listdir(compressed_train))\n",
    "\n",
    "for video in video_files:\n",
    "    video_path = os.path.join(compressed_train, video)\n",
    "    cap = cv2.VideoCapture(video_path)\n",
    "    frame_count = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))\n",
    "    \n",
    "    frameNr = 0\n",
    "    \n",
    "    while True:\n",
    "        ret, frame = cap.read()\n",
    "        \n",
    "        if ret:\n",
    "            if frameNr % 250 == 0:\n",
    "                # Save every 250th frame\n",
    "                cv2.imwrite(f'images_train_compressed_ALL/train_{nr}_{video}_frame{frameNr}.png', frame)\n",
    "                nr += 1\n",
    "        else:\n",
    "            break\n",
    "        \n",
    "        frameNr += 1\n",
    "    \n",
    "    cap.release()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Dataset with REFERENCE-frames, png format neccessary.\n",
    "\n",
    "nr = 1\n",
    "\n",
    "video_files = sorted(os.listdir(train))\n",
    "\n",
    "for video in video_files:\n",
    "    video_path = os.path.join(train, video)\n",
    "    cap = cv2.VideoCapture(video_path)\n",
    "    frame_count = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))\n",
    "    \n",
    "    frameNr = 0\n",
    "    \n",
    "    while True:\n",
    "        ret, frame = cap.read()\n",
    "        \n",
    "        if ret:\n",
    "            if frameNr % 250 == 0:\n",
    "                # Save every 250th frame\n",
    "                cv2.imwrite(f'images_train/train_{nr}_{video}_frame{frameNr}.png', frame)\n",
    "                nr += 1\n",
    "        else:\n",
    "            break\n",
    "        \n",
    "        frameNr += 1\n",
    "    \n",
    "    cap.release()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Dataset with compressed frames TEST SET, png format neccessary.\n",
    "\n",
    "compressed_test = \"compressed_TEST_videos_ALL\"\n",
    "nr = 1\n",
    "\n",
    "video_files = sorted(os.listdir(compressed_test))\n",
    "\n",
    "for video in video_files:\n",
    "    video_path = os.path.join(compressed_test, video)\n",
    "    cap = cv2.VideoCapture(video_path)\n",
    "    frame_count = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))\n",
    "    \n",
    "    frameNr = 0\n",
    "    \n",
    "    while True:\n",
    "        ret, frame = cap.read()\n",
    "        \n",
    "        if ret:\n",
    "            if frameNr % 250 == 0:\n",
    "                # Save every 250th frame\n",
    "                cv2.imwrite(f'images_TEST_compressed_ALL/TEST_{nr}_{video}_frame{frameNr}.png', frame)\n",
    "                nr += 1\n",
    "        else:\n",
    "            break\n",
    "        \n",
    "        frameNr += 1\n",
    "    \n",
    "    cap.release()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Dataset with REFERENCE-frames TEST SET, png format neccessary.\n",
    "\n",
    "nr = 1\n",
    "\n",
    "video_files = sorted(os.listdir(test))\n",
    "\n",
    "for video in video_files:\n",
    "    video_path = os.path.join(test, video)\n",
    "    cap = cv2.VideoCapture(video_path)\n",
    "    frame_count = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))\n",
    "    \n",
    "    frameNr = 0\n",
    "    \n",
    "    while True:\n",
    "        ret, frame = cap.read()\n",
    "        \n",
    "        if ret:\n",
    "            if frameNr % 250 == 0:\n",
    "                # Save every 250th frame\n",
    "                cv2.imwrite(f'images_TEST/TEST_{nr}_{video}_frame{frameNr}.png', frame)\n",
    "                nr += 1\n",
    "        else:\n",
    "            break\n",
    "        \n",
    "        frameNr += 1\n",
    "    \n",
    "    cap.release()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Reference frames CROP, training set.\n",
    "\n",
    "input_folder = \"images_train\"\n",
    "output_folder = \"images_train_crop_png\"\n",
    "crop_size = (224, 224)   # Default input size of DenseNet-121.\n",
    "\n",
    "image_files = [f for f in os.listdir(input_folder) if f.endswith(('.png'))]\n",
    "\n",
    "for filename in image_files:\n",
    "    im_path = os.path.join(input_folder, filename)\n",
    "    im = Image.open(im_path)\n",
    "    width, height = im.size\n",
    "\n",
    "    num_chunks_x = width // crop_size[0]         # number of columns\n",
    "    num_chunks_y = height // crop_size[1]        # number of rows\n",
    "\n",
    "    for i in range(0, num_chunks_x):\n",
    "        for j in range(0, num_chunks_y):\n",
    "            \n",
    "            left = i * crop_size[0]\n",
    "            top = j * crop_size[1]\n",
    "            right = left + crop_size[0]\n",
    "            bottom = top + crop_size[1]\n",
    "\n",
    "            cropped = im.crop((left, top, right, bottom))\n",
    "            \n",
    "            location_info = \"\"\n",
    "            if i == 0:\n",
    "                location_info += \"left_\"\n",
    "            elif i == num_chunks_x - 1:\n",
    "                location_info += \"right_\"\n",
    "            if j == 0:\n",
    "                location_info += \"top\"\n",
    "            elif j == num_chunks_y - 1:\n",
    "                location_info += \"bottom\"\n",
    "                \n",
    "            if left <= 960 <= right and top <= 540 <= bottom:   # center of the frame\n",
    "                location_info += \"center_\"\n",
    "                \n",
    "            \n",
    "            pos_x = left\n",
    "            pos_y = top\n",
    "            \n",
    "            output_filename = f\"{os.path.splitext(filename)[0]}_{location_info}_start_{pos_x}_{pos_y}.png\"\n",
    "            output_path = os.path.join(output_folder, output_filename)\n",
    "            cropped.save(output_path)\n",
    "\n",
    "    im.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compressed frames CROP, training set.\n",
    "\n",
    "input_folder = \"images_train_compressed_ALL\"\n",
    "output_folder = \"images_train_comp_crop_ALL\"\n",
    "crop_size = (224, 224)   \n",
    "\n",
    "image_files = [f for f in os.listdir(input_folder) if f.endswith(('.jpg', '.png'))]\n",
    "\n",
    "for filename in image_files:\n",
    "    im_path = os.path.join(input_folder, filename)\n",
    "    im = Image.open(im_path)\n",
    "    width, height = im.size\n",
    "\n",
    "    num_chunks_x = width // crop_size[0]        \n",
    "    num_chunks_y = height // crop_size[1]        \n",
    "\n",
    "    for i in range(0, num_chunks_x):\n",
    "        for j in range(0, num_chunks_y):\n",
    "            \n",
    "            left = i * crop_size[0]\n",
    "            top = j * crop_size[1]\n",
    "            right = left + crop_size[0]\n",
    "            bottom = top + crop_size[1]\n",
    "\n",
    "            cropped = im.crop((left, top, right, bottom))\n",
    "            \n",
    "            location_info = \"\"\n",
    "            if i == 0:\n",
    "                location_info += \"left_\"\n",
    "            elif i == num_chunks_x - 1:\n",
    "                location_info += \"right_\"\n",
    "            if j == 0:\n",
    "                location_info += \"top\"\n",
    "            elif j == num_chunks_y - 1:\n",
    "                location_info += \"bottom\"\n",
    "                \n",
    "            if left <= 960 <= right and top <= 540 <= bottom:   \n",
    "                location_info += \"center_\"\n",
    "                \n",
    "            \n",
    "            pos_x = left\n",
    "            pos_y = top\n",
    "            \n",
    "            output_filename = f\"{os.path.splitext(filename)[0]}_{location_info}_start_{pos_x}_{pos_y}.png\"\n",
    "            output_path = os.path.join(output_folder, output_filename)\n",
    "            cropped.save(output_path)\n",
    "\n",
    "    im.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# compressed frames CROP, test data.\n",
    "\n",
    "input_folder = \"images_TEST_compressed_ALL\"\n",
    "output_folder = \"images_TEST_comp_crop_png_ALL\"\n",
    "crop_size = (224, 224)   \n",
    "\n",
    "image_files = [f for f in os.listdir(input_folder) if f.endswith(('.jpg', '.png'))]\n",
    "\n",
    "for filename in image_files:\n",
    "    im_path = os.path.join(input_folder, filename)\n",
    "    im = Image.open(im_path)\n",
    "    width, height = im.size\n",
    "\n",
    "    num_chunks_x = width // crop_size[0]        \n",
    "    num_chunks_y = height // crop_size[1]        \n",
    "\n",
    "    for i in range(0, num_chunks_x):\n",
    "        for j in range(0, num_chunks_y):\n",
    "            \n",
    "            left = i * crop_size[0]\n",
    "            top = j * crop_size[1]\n",
    "            right = left + crop_size[0]\n",
    "            bottom = top + crop_size[1]\n",
    "\n",
    "            cropped = im.crop((left, top, right, bottom))\n",
    "            \n",
    "            location_info = \"\"\n",
    "            if i == 0:\n",
    "                location_info += \"left_\"\n",
    "            elif i == num_chunks_x - 1:\n",
    "                location_info += \"right_\"\n",
    "            if j == 0:\n",
    "                location_info += \"top\"\n",
    "            elif j == num_chunks_y - 1:\n",
    "                location_info += \"bottom\"\n",
    "                \n",
    "            if left <= 960 <= right and top <= 540 <= bottom:   \n",
    "                location_info += \"center_\"\n",
    "                \n",
    "            \n",
    "            pos_x = left\n",
    "            pos_y = top\n",
    "            \n",
    "            output_filename = f\"{os.path.splitext(filename)[0]}_{location_info}_start_{pos_x}_{pos_y}.png\"\n",
    "            output_path = os.path.join(output_folder, output_filename)\n",
    "            cropped.save(output_path)\n",
    "\n",
    "\n",
    "    im.close()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Reference frames CROP, test data.\n",
    "\n",
    "input_folder = \"images_TEST\"\n",
    "output_folder = \"images_TEST_crop_png\"\n",
    "crop_size = (224, 224)   \n",
    "\n",
    "image_files = [f for f in os.listdir(input_folder) if f.endswith(('.png'))]\n",
    "\n",
    "for filename in image_files:\n",
    "    im_path = os.path.join(input_folder, filename)\n",
    "    im = Image.open(im_path)\n",
    "    width, height = im.size\n",
    "\n",
    "    num_chunks_x = width // crop_size[0]         \n",
    "    num_chunks_y = height // crop_size[1]        \n",
    "\n",
    "    for i in range(0, num_chunks_x):\n",
    "        for j in range(0, num_chunks_y):\n",
    "            \n",
    "            left = i * crop_size[0]\n",
    "            top = j * crop_size[1]\n",
    "            right = left + crop_size[0]\n",
    "            bottom = top + crop_size[1]\n",
    "\n",
    "            cropped = im.crop((left, top, right, bottom))\n",
    "            \n",
    "            location_info = \"\"\n",
    "            if i == 0:\n",
    "                location_info += \"left_\"\n",
    "            elif i == num_chunks_x - 1:\n",
    "                location_info += \"right_\"\n",
    "            if j == 0:\n",
    "                location_info += \"top\"\n",
    "            elif j == num_chunks_y - 1:\n",
    "                location_info += \"bottom\"\n",
    "                \n",
    "            if left <= 960 <= right and top <= 540 <= bottom:   \n",
    "                location_info += \"center_\"\n",
    "                \n",
    "            \n",
    "            pos_x = left\n",
    "            pos_y = top\n",
    "            \n",
    "            output_filename = f\"{os.path.splitext(filename)[0]}_{location_info}_start_{pos_x}_{pos_y}.png\"\n",
    "            output_path = os.path.join(output_folder, output_filename)\n",
    "            cropped.save(output_path)\n",
    "\n",
    "\n",
    "    im.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Check if lengths are correct.\n",
    "\n",
    "print(len(os.listdir(\"images_train_comp_crop_png\")))\n",
    "print(len(os.listdir(\"images_train_crop_png\")))\n",
    "\n",
    "print(len(os.listdir(\"images_TEST_crop_png\")))\n",
    "print(len(os.listdir(\"images_TEST_comp_crop_png\")))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def extract_details_comp(path):\n",
    "    \"\"\"\n",
    "    Extracts movie name, frame number, and crop details from the path.\n",
    "    \"\"\"\n",
    "    components = path.split(\"_\")\n",
    "    compressed = components[7]\n",
    "    movie_name = components[8]\n",
    "    frame_number = components[9]\n",
    "    crop_details = \"_\".join(components[-2:])\n",
    "    return compressed, movie_name, frame_number, crop_details\n",
    "\n",
    "\n",
    "compressed_crops_23 = []\n",
    "compressed_crops_37 = []\n",
    "compressed_crops_51 = []\n",
    "for image in os.listdir(\"images_train_comp_crop_ALL\"):\n",
    "    im_path = os.path.join(\"images_train_comp_crop_ALL\", image)\n",
    "    \n",
    "    components = im_path.split(\"_\")\n",
    "    \n",
    "print(components)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "reference_crops = []\n",
    "compressed_crops_23 = []\n",
    "compressed_crops_50 = []\n",
    "compressed_crops_51 = []\n",
    "\n",
    "for image in os.listdir(\"images_train_crop_png\"):\n",
    "    im_path = os.path.join(\"images_train_crop_png\", image)\n",
    "    \n",
    "    reference_crops.append(im_path)\n",
    "\n",
    "for image in os.listdir(\"images_train_comp_crop_ALL\"):\n",
    "    im_path = os.path.join(\"images_train_comp_crop_ALL\", image)\n",
    "    \n",
    "    components = im_path.split(\"_\")\n",
    "    compressed = components[7]\n",
    "     \n",
    "    if compressed == \"23\":\n",
    "        compressed_crops_23.append(im_path)\n",
    "        \n",
    "    elif compressed == \"50\":\n",
    "        compressed_crops_50.append(im_path)\n",
    "        \n",
    "    elif compressed == \"51\":\n",
    "        compressed_crops_51.append(im_path)\n",
    "        \n",
    "\n",
    "def extract_details_ref(path):\n",
    "    components = path.split(\"_\")\n",
    "    movie_name = components[5]\n",
    "    frame_number = components[6]\n",
    "    crop_details = \"_\".join(components[-2:])\n",
    "    return movie_name, frame_number, crop_details\n",
    "\n",
    "\n",
    "def extract_details_comp(path):\n",
    "    components = path.split(\"_\")\n",
    "    movie_name = components[8]\n",
    "    frame_number = components[9]\n",
    "    crop_details = \"_\".join(components[-2:])\n",
    "    return movie_name, frame_number, crop_details\n",
    "\n",
    "\n",
    "def sort_by_details_ref(path):\n",
    "    movie_name, frame_number, crop_details = extract_details_ref(path)\n",
    "    return movie_name, frame_number, crop_details\n",
    "\n",
    "\n",
    "def sort_by_details_comp(path):\n",
    "    movie_name, frame_number, crop_details = extract_details_comp(path)\n",
    "    return movie_name, frame_number, crop_details\n",
    "\n",
    "\n",
    "reference_sorted = sorted(reference_crops, key=sort_by_details_ref)\n",
    "compressed_sorted_23 = sorted(compressed_crops_23, key=sort_by_details_comp)\n",
    "compressed_sorted_50 = sorted(compressed_crops_37, key=sort_by_details_comp)\n",
    "compressed_sorted_51 = sorted(compressed_crops_51, key=sort_by_details_comp)\n",
    "\n",
    "print(reference_sorted[31])\n",
    "print(compressed_sorted_51[31])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# VMAF calculation.\n",
    "\n",
    "output_file = \"vmaf_values_train_ALL.csv\" \n",
    "compression_levels = [23, 50, 51]\n",
    "\n",
    "for comp in compression_levels:\n",
    "    if comp == 23:\n",
    "        if len(reference_sorted) != len(compressed_sorted_23):\n",
    "            print(\"Error: The number of files in the two folders is not the same.\")\n",
    "            exit()\n",
    "\n",
    "        vmaf_scores = []\n",
    "        count = 0\n",
    "            \n",
    "        for filename1, filename2 in zip(reference_sorted, compressed_sorted_23):\n",
    "                image_path_ref = filename1\n",
    "                image_path_dist = filename2\n",
    "                \n",
    "                # image_path_dist is compresssed images, image_path_ref is reference images.\n",
    "                command = ['/opt/homebrew/bin/ffmpeg', '-i', image_path_dist, '-i', image_path_ref, '-lavfi', 'libvmaf', '-f', 'null', '-']\n",
    "                output = subprocess.run(command, capture_output=True, text=True)\n",
    "                \n",
    "                pattern = re.compile(r'VMAF score: (\\d+\\.\\d+)')  # (\\d+\\.\\d+) matches the VMAF score\n",
    "                match = pattern.search(str(output))\n",
    "                \n",
    "                if match:\n",
    "                    vmaf_score = match.group(1)\n",
    "                    vmaf_scores.append([vmaf_score, image_path_dist])\n",
    "                    count +=1\n",
    "                \n",
    "        print(count)\n",
    "        \n",
    "        \n",
    "    elif comp == 50:\n",
    "        if len(reference_sorted) != len(compressed_sorted_50):\n",
    "            print(\"Error: The number of files in the two folders is not the same.\")\n",
    "            exit()\n",
    "            \n",
    "        for filename1, filename2 in zip(reference_sorted, compressed_sorted_50):\n",
    "                image_path_ref = filename1\n",
    "                image_path_dist = filename2\n",
    "                \n",
    "                command = ['/opt/homebrew/bin/ffmpeg', '-i', image_path_dist, '-i', image_path_ref, '-lavfi', 'libvmaf', '-f', 'null', '-']\n",
    "                output = subprocess.run(command, capture_output=True, text=True)\n",
    "                \n",
    "                pattern = re.compile(r'VMAF score: (\\d+\\.\\d+)')  # (\\d+\\.\\d+) matches the VMAF score\n",
    "                match = pattern.search(str(output))\n",
    "                \n",
    "                if match:\n",
    "                    vmaf_score = match.group(1)\n",
    "                    vmaf_scores.append([vmaf_score, image_path_dist])\n",
    "                    count +=1\n",
    "                \n",
    "        print(count)\n",
    "        \n",
    "    elif comp == 51:\n",
    "        if len(reference_sorted) != len(compressed_sorted_51):\n",
    "            print(\"Error: The number of files in the two folders is not the same.\")\n",
    "            exit()\n",
    "\n",
    "            \n",
    "        for filename1, filename2 in zip(reference_sorted, compressed_sorted_51):\n",
    "                image_path_ref = filename1\n",
    "                image_path_dist = filename2\n",
    "                \n",
    "                command = ['/opt/homebrew/bin/ffmpeg', '-i', image_path_dist, '-i', image_path_ref, '-lavfi', 'libvmaf', '-f', 'null', '-']\n",
    "                output = subprocess.run(command, capture_output=True, text=True)\n",
    "                \n",
    "                pattern = re.compile(r'VMAF score: (\\d+\\.\\d+)')  # (\\d+\\.\\d+) matches the VMAF score\n",
    "                match = pattern.search(str(output))\n",
    "                \n",
    "                if match:\n",
    "                    vmaf_score = match.group(1)\n",
    "                    vmaf_scores.append([vmaf_score, image_path_dist])\n",
    "                    count +=1\n",
    "                \n",
    "        print(count)\n",
    "            \n",
    "            \n",
    "with open(output_file, 'w', newline='') as csvfile:     # Write VMAF and corresponding VMAF value to file.\n",
    "    csv_writer = csv.writer(csvfile)\n",
    "    csv_writer.writerow(['VMAF Score', 'Filename'])\n",
    "    csv_writer.writerows(vmaf_scores)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "reference_frames = []\n",
    "compressed_frames_23 = []\n",
    "compressed_frames_50 = []\n",
    "compressed_frames_51 = []\n",
    "\n",
    "for image in os.listdir(\"images_train\"):\n",
    "    im_path = os.path.join(\"images_train\", image)\n",
    "    \n",
    "    reference_frames.append(im_path)\n",
    "    \n",
    "for image in os.listdir(\"images_train_compressed_ALL\"):\n",
    "    im_path = os.path.join(\"images_train_compressed_ALL\", image)\n",
    "    \n",
    "    components = im_path.split(\"_\")\n",
    "    compressed = components[6]\n",
    "     \n",
    "    if compressed == \"23\":\n",
    "        compressed_frames_23.append(im_path)\n",
    "        \n",
    "    elif compressed == \"50\":\n",
    "        compressed_frames_50.append(im_path)\n",
    "        \n",
    "    elif compressed == \"51\":\n",
    "        compressed_frames_51.append(im_path)\n",
    "        \n",
    "\n",
    "def sort_by_movie_ref(path):\n",
    "    return path.split(\"_\")[3]    # film trailer name\n",
    "   \n",
    "def sort_by_movie_comp(path):\n",
    "    return path.split(\"_\")[7]     # compression level\n",
    "\n",
    "\n",
    "def sort_by_frame(path):\n",
    "    return path.split(\"_\")[-1]       \n",
    "\n",
    "\n",
    "reference_sorted = sorted(reference_frames, key=sort_by_movie_ref)\n",
    "compressed_sorted_23 = sorted(compressed_frames_23, key=sort_by_movie_comp)\n",
    "compressed_sorted_50 = sorted(compressed_frames_50, key=sort_by_movie_comp)\n",
    "compressed_sorted_51 = sorted(compressed_frames_51, key=sort_by_movie_comp)\n",
    "\n",
    "reference_sorted_by_frames = sorted(reference_sorted, key=sort_by_frame)\n",
    "compressed_sorted_by_frames_51 = sorted(compressed_sorted_51, key=sort_by_frame)\n",
    "compressed_sorted_by_frames_50 = sorted(compressed_sorted_50, key=sort_by_frame)\n",
    "compressed_sorted_by_frames_23 = sorted(compressed_sorted_23, key=sort_by_frame)\n",
    "\n",
    "# print(compressed_sorted)\n",
    "\n",
    "print(reference_sorted_by_frames[65])\n",
    "print(compressed_sorted_by_frames_50[65])\n",
    "\n",
    "plt.imshow(plt.imread(reference_sorted_by_frames[250]))  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# VMAF on full frames, for analysis.\n",
    "output_file2 = \"vmaf_on_full_frames_train_ALL.csv\" \n",
    "\n",
    "\n",
    "for comp in compression_levels:\n",
    "    if comp == 23:\n",
    "        if len(reference_sorted_by_frames) != len(compressed_sorted_by_frames_23):\n",
    "            print(\"Error: The number of files in the two folders is not the same.\")\n",
    "            exit()\n",
    "\n",
    "        vmaf_scores = []\n",
    "        count = 0\n",
    "            \n",
    "        for filename1, filename2 in zip(reference_sorted_by_frames, compressed_sorted_by_frames_23):\n",
    "                image_path_ref = filename1\n",
    "                image_path_dist = filename2\n",
    "                \n",
    "                command = ['/opt/homebrew/bin/ffmpeg', '-i', image_path_dist, '-i', image_path_ref, '-lavfi', 'libvmaf', '-f', 'null', '-']\n",
    "                output = subprocess.run(command, capture_output=True, text=True)\n",
    "                \n",
    "                pattern = re.compile(r'VMAF score: (\\d+\\.\\d+)')  # (\\d+\\.\\d+) matches the VMAF score\n",
    "                match = pattern.search(str(output))\n",
    "                \n",
    "                if match:\n",
    "                    vmaf_score = match.group(1)\n",
    "                    vmaf_scores.append([vmaf_score, image_path_dist])\n",
    "                    count +=1\n",
    "                \n",
    "        print(count)\n",
    "        \n",
    "        \n",
    "    elif comp == 50:\n",
    "        if len(reference_sorted_by_frames) != len(compressed_sorted_by_frames_50):\n",
    "            print(\"Error: The number of files in the two folders is not the same.\")\n",
    "            exit()\n",
    "            \n",
    "        for filename1, filename2 in zip(reference_sorted_by_frames, compressed_sorted_by_frames_50):\n",
    "                image_path_ref = filename1\n",
    "                image_path_dist = filename2\n",
    "                \n",
    "                command = ['/opt/homebrew/bin/ffmpeg', '-i', image_path_dist, '-i', image_path_ref, '-lavfi', 'libvmaf', '-f', 'null', '-']\n",
    "                output = subprocess.run(command, capture_output=True, text=True)\n",
    "                \n",
    "                pattern = re.compile(r'VMAF score: (\\d+\\.\\d+)')  # (\\d+\\.\\d+) matches the VMAF score\n",
    "                match = pattern.search(str(output))\n",
    "                \n",
    "                if match:\n",
    "                    vmaf_score = match.group(1)\n",
    "                    vmaf_scores.append([vmaf_score, image_path_dist])\n",
    "                    count +=1\n",
    "                \n",
    "        print(count)\n",
    "        \n",
    "    elif comp == 51:\n",
    "        if len(reference_sorted_by_frames) != len(compressed_sorted_by_frames_51):\n",
    "            print(\"Error: The number of files in the two folders is not the same.\")\n",
    "            exit()\n",
    "\n",
    "            \n",
    "        for filename1, filename2 in zip(reference_sorted_by_frames, compressed_sorted_by_frames_51):\n",
    "                image_path_ref = filename1\n",
    "                image_path_dist = filename2\n",
    "                \n",
    "                command = ['/opt/homebrew/bin/ffmpeg', '-i', image_path_dist, '-i', image_path_ref, '-lavfi', 'libvmaf', '-f', 'null', '-']\n",
    "                output = subprocess.run(command, capture_output=True, text=True)\n",
    "                \n",
    "                pattern = re.compile(r'VMAF score: (\\d+\\.\\d+)')  # (\\d+\\.\\d+) matches the VMAF score\n",
    "                match = pattern.search(str(output))\n",
    "                \n",
    "                if match:\n",
    "                    vmaf_score = match.group(1)\n",
    "                    vmaf_scores.append([vmaf_score, image_path_dist])\n",
    "                    count +=1\n",
    "                \n",
    "        print(count)\n",
    "            \n",
    "            \n",
    "with open(output_file2, 'w', newline='') as csvfile:\n",
    "    csv_writer = csv.writer(csvfile)\n",
    "    csv_writer.writerow(['VMAF Score', 'Filename'])\n",
    "    csv_writer.writerows(vmaf_scores)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "Distribution of VMAF for analysis\n",
    "\"\"\"\n",
    "\n",
    "file = pd.read_csv(\"vmaf_values_train_ALL.csv\")\n",
    "\n",
    "file2 = pd.read_csv(\"vmaf_on_full_frames_train_ALL.csv\")\n",
    "\n",
    "fig, axs = plt.subplots(1, 2, figsize=(15, 6))\n",
    "\n",
    "axs[0].hist(file.iloc[:, 0], edgecolor=\"k\", color=\"c\")\n",
    "axs[0].grid(alpha=0.5)\n",
    "axs[0].set_xlabel(\"VMAF values on cropped images\")\n",
    "axs[0].set_xticks(range(0, 101, 10))\n",
    "axs[0].grid(alpha=0.5)\n",
    "axs[0].set_title(\"VMAF distribution in cropped frames\")\n",
    "\n",
    "axs[1].hist(file2.iloc[:, 0], edgecolor=\"k\", color=\"darkblue\", alpha=0.75)\n",
    "axs[1].set_xlabel(\"VMAF values on full-size frames\")\n",
    "axs[1].set_ylabel('Frequency')\n",
    "axs[1].set_xticks(range(0, 101, 10))\n",
    "axs[1].grid(alpha=0.5)\n",
    "axs[1].set_title(\"VMAF distribution in full frames\")\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "csv_file_path = \"vmaf_values_train_ALL.csv\"\n",
    "\n",
    "positions = []\n",
    "left = []\n",
    "right = []\n",
    "bottom = []\n",
    "top = []\n",
    "center = []\n",
    "corners = []\n",
    "\n",
    "with open(csv_file_path, newline='') as csvfile:\n",
    "    csv_reader = csv.reader(csvfile)\n",
    "    for row in csv_reader:\n",
    "        vmaf_score = row[0]\n",
    "        filename = row[1]\n",
    "        \n",
    "        position_info = filename.split('/')[-1].split('_')[6:-1]\n",
    "        position = [vmaf_score, '_'.join(position_info)]\n",
    "        positions.append(position)\n",
    "        \n",
    "        \n",
    "    for pos in positions:\n",
    "        if \"center\" in str(pos[1]):\n",
    "            center.append(pos)\n",
    "            \n",
    "        elif \"top\" in str(pos[1]):\n",
    "            top.append(pos)\n",
    "            \n",
    "            if \"left\" or \"right\" in str(pos[1]):\n",
    "                corners.append(pos)\n",
    "            \n",
    "        elif \"bottom\" in str(pos[1]):\n",
    "            bottom.append(pos)\n",
    "            \n",
    "            if \"left\" or \"right\" in str(pos[1]):\n",
    "                corners.append(pos)\n",
    "        \n",
    "            \n",
    "        elif \"right\" in str(pos[1]):\n",
    "            right.append(pos)\n",
    "            \n",
    "        elif \"left\" in str(pos[1]):\n",
    "            left.append(pos)\n",
    "            \n",
    "\n",
    "vmaf_left = [float(item[0]) for item in left]\n",
    "vmaf_right = [float(item[0]) for item in right]\n",
    "vmaf_center = [float(item[0]) for item in center]\n",
    "\n",
    "\n",
    "fig, axs = plt.subplots(1, 3, figsize=(20, 6))\n",
    "axs[0].hist(sorted(vmaf_left), edgecolor=\"k\", alpha=0.75)\n",
    "axs[0].set_xlabel('VMAF Score')\n",
    "axs[0].set_ylabel('Frequency')\n",
    "axs[0].set_title('Frequency Distribution of VMAF Scores in left-aligned cropped images')\n",
    "axs[0].set_xticks(range(0, 101, 10))\n",
    "axs[0].grid(alpha=0.5)\n",
    "\n",
    "axs[1].hist(sorted(vmaf_center), edgecolor=\"k\", color=\"g\", alpha=0.75)\n",
    "axs[1].set_xlabel('VMAF Score')\n",
    "axs[1].set_ylabel('Frequency')\n",
    "axs[1].set_title('Frequency Distribution of VMAF Scores in center-aligned cropped images')\n",
    "axs[1].set_xticks(range(0, 101, 10))\n",
    "axs[1].grid(alpha=0.5)\n",
    "\n",
    "axs[2].hist(sorted(vmaf_right), edgecolor=\"k\", color=\"r\", alpha=0.75)\n",
    "axs[2].set_xlabel('VMAF Score')\n",
    "axs[2].set_ylabel('Frequency')\n",
    "axs[2].set_title('Frequency Distribution of VMAF Scores in right-aligned cropped images')\n",
    "axs[2].set_xticks(range(0, 101, 10))\n",
    "axs[2].grid(alpha=0.5)\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "vmaf_corners = [float(item[0]) for item in corners]\n",
    "\n",
    "fig, axs = plt.subplots(1, 2, figsize=(15, 6))\n",
    "\n",
    "axs[0].hist(sorted(vmaf_corners), color = \"orange\", edgecolor=\"k\")\n",
    "axs[0].set_xlabel('VMAF Score')\n",
    "axs[0].set_ylabel('Frequency')\n",
    "axs[0].set_title('Frequency Distribution of VMAF Scores in corners')    # corners include left bottom, left top, right bottom, right top\n",
    "axs[0].set_xticks(range(0, 101, 10))\n",
    "axs[0].grid(alpha=0.5)\n",
    "\n",
    "axs[1].hist(sorted(vmaf_center), color = \"green\", edgecolor=\"k\", alpha=0.75)\n",
    "axs[1].set_xlabel('VMAF Score')\n",
    "axs[1].set_ylabel('Frequency')\n",
    "axs[1].set_title('Frequency Distribution of VMAF Scores in centers')\n",
    "axs[1].set_xticks(range(0, 101, 10))\n",
    "axs[1].grid(alpha=0.5)\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(len(os.listdir(\"images_train_comp_crop_ALL\")))\n",
    "df = pd.read_csv('vmaf_values_train_ALL.csv')\n",
    "print(df.shape[0])      "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "for image in os.listdir(\"images_TEST_comp_crop_png_ALL\"):\n",
    "    im_path = os.path.join(\"images_TEST_comp_crop_png_ALL\", image)\n",
    "    components = im_path.split(\"_\")\n",
    "print(components[-2:])  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "reference_crops_test = []\n",
    "compressed_crops_test_23 = []\n",
    "compressed_crops_test_50 = []\n",
    "compressed_crops_test_51 = []\n",
    "\n",
    "def extract_details_ref(path):\n",
    "    components = path.split(\"_\")\n",
    "    movie_name = components[5]\n",
    "    frame_number = components[6]\n",
    "    crop_details = \"_\".join(components[-2:])\n",
    "    return movie_name, frame_number, crop_details\n",
    "\n",
    "\n",
    "def extract_details_comp(path):\n",
    "    components = path.split(\"_\")\n",
    "    movie_name = components[9]\n",
    "    frame_number = components[10]\n",
    "    crop_details = \"_\".join(components[-2:])\n",
    "    return movie_name, frame_number, crop_details\n",
    "\n",
    "\n",
    "def sort_by_details_ref(path):\n",
    "    movie_name, frame_number, crop_details = extract_details_ref(path)\n",
    "    return movie_name, frame_number, crop_details\n",
    "\n",
    "\n",
    "def sort_by_details_comp(path):\n",
    "    movie_name, frame_number, crop_details = extract_details_comp(path)\n",
    "    return movie_name, frame_number, crop_details\n",
    "\n",
    "\n",
    "for image in os.listdir(\"images_TEST_crop_png\"):\n",
    "    im_path = os.path.join(\"images_TEST_crop_png\", image)\n",
    "    \n",
    "    reference_crops_test.append(im_path)\n",
    "\n",
    "for image in os.listdir(\"images_TEST_comp_crop_png_ALL\"):\n",
    "    im_path = os.path.join(\"images_TEST_comp_crop_png_ALL\", image)\n",
    "    components = im_path.split(\"_\")\n",
    "    \n",
    "    if components[8] == \"23\":\n",
    "        compressed_crops_test_23.append(im_path)\n",
    "        \n",
    "    elif components[8] == \"50\":\n",
    "        compressed_crops_test_50.append(im_path)\n",
    "        \n",
    "    elif components[8] == \"51\":\n",
    "        compressed_crops_test_51.append(im_path)\n",
    "\n",
    "\n",
    "reference_sorted_test = sorted(reference_crops_test, key=sort_by_details_ref)\n",
    "compressed_sorted_test_23 = sorted(compressed_crops_test_23, key=sort_by_details_comp)\n",
    "compressed_sorted_test_50 = sorted(compressed_crops_test_50, key=sort_by_details_comp)\n",
    "compressed_sorted_test_51 = sorted(compressed_crops_test_51, key=sort_by_details_comp)\n",
    "\n",
    "print(reference_sorted_test[265])\n",
    "print(compressed_sorted_test_51[265])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# VMAF values TEST-set.\n",
    "output_file_test = \"vmaf_values_TEST_ALL.csv\" \n",
    "compression_levels = [23, 50, 51]\n",
    "\n",
    "for comp in compression_levels:\n",
    "    if comp == 23:\n",
    "        if len(reference_sorted_test) != len(compressed_sorted_test_23):\n",
    "            print(\"Error: The number of files in the two folders is not the same.\")\n",
    "            exit()\n",
    "\n",
    "        vmaf_scores = []\n",
    "        count = 0\n",
    "            \n",
    "        for filename1, filename2 in zip(reference_sorted_test, compressed_sorted_test_23):\n",
    "                image_path_ref = filename1\n",
    "                image_path_dist = filename2\n",
    "                \n",
    "                command = ['/opt/homebrew/bin/ffmpeg', '-i', image_path_dist, '-i', image_path_ref, '-lavfi', 'libvmaf', '-f', 'null', '-']\n",
    "                output = subprocess.run(command, capture_output=True, text=True)\n",
    "                \n",
    "                pattern = re.compile(r'VMAF score: (\\d+\\.\\d+)')  # (\\d+\\.\\d+) matches the VMAF score\n",
    "                match = pattern.search(str(output))\n",
    "                \n",
    "                if match:\n",
    "                    vmaf_score = match.group(1)\n",
    "                    vmaf_scores.append([vmaf_score, image_path_dist])\n",
    "                    count +=1\n",
    "                \n",
    "        print(count)\n",
    "        \n",
    "        \n",
    "    elif comp == 50:\n",
    "        if len(reference_sorted_test) != len(compressed_sorted_test_50):\n",
    "            print(\"Error: The number of files in the two folders is not the same.\")\n",
    "            exit()\n",
    "            \n",
    "        for filename1, filename2 in zip(reference_sorted_test, compressed_sorted_test_50):\n",
    "                image_path_ref = filename1\n",
    "                image_path_dist = filename2\n",
    "                \n",
    "                command = ['/opt/homebrew/bin/ffmpeg', '-i', image_path_dist, '-i', image_path_ref, '-lavfi', 'libvmaf', '-f', 'null', '-']\n",
    "                output = subprocess.run(command, capture_output=True, text=True)\n",
    "                \n",
    "                pattern = re.compile(r'VMAF score: (\\d+\\.\\d+)')  # (\\d+\\.\\d+) matches the VMAF score\n",
    "                match = pattern.search(str(output))\n",
    "                \n",
    "                if match:\n",
    "                    vmaf_score = match.group(1)\n",
    "                    vmaf_scores.append([vmaf_score, image_path_dist])\n",
    "                    count +=1\n",
    "                \n",
    "        print(count)\n",
    "        \n",
    "    elif comp == 51:\n",
    "        if len(reference_sorted_test) != len(compressed_sorted_test_51):\n",
    "            print(\"Error: The number of files in the two folders is not the same.\")\n",
    "            exit()\n",
    "\n",
    "            \n",
    "        for filename1, filename2 in zip(reference_sorted_test, compressed_sorted_test_51):\n",
    "                image_path_ref = filename1\n",
    "                image_path_dist = filename2\n",
    "                \n",
    "                command = ['/opt/homebrew/bin/ffmpeg', '-i', image_path_dist, '-i', image_path_ref, '-lavfi', 'libvmaf', '-f', 'null', '-']\n",
    "                output = subprocess.run(command, capture_output=True, text=True)\n",
    "                \n",
    "                pattern = re.compile(r'VMAF score: (\\d+\\.\\d+)')  # (\\d+\\.\\d+) matches the VMAF score\n",
    "                match = pattern.search(str(output))\n",
    "                \n",
    "                if match:\n",
    "                    vmaf_score = match.group(1)\n",
    "                    vmaf_scores.append([vmaf_score, image_path_dist])\n",
    "                    count +=1\n",
    "                \n",
    "        print(count)\n",
    "             \n",
    "with open(output_file_test, 'w', newline='') as csvfile:\n",
    "    csv_writer = csv.writer(csvfile)\n",
    "    csv_writer.writerow(['VMAF Score', 'Filename'])\n",
    "    csv_writer.writerows(vmaf_scores)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(pd.read_csv(\"vmaf_values_train_ALL.csv\").iloc[9000, 0])\n",
    "print(os.path.join(\"images_train_comp_crop_ALL\", pd.read_csv(\"vmaf_values_train_ALL.csv\").iloc[9000, 1]))\n",
    "plt.imshow(mpimg.imread(os.path.join(\"images_train_comp_crop_ALL\", pd.read_csv(\"vmaf_values_train_ALL.csv\").iloc[9000, 1])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'\\nTo train DenseNet on train data:\\n1) loop over epochs (passes in dataset)\\n2) loop over minibatches\\n3) forward pass\\n4) compute loss\\n5) backward pass\\n6) upgrade parameters with optimizer (Adam)\\n\\n'"
      ]
     },
     "execution_count": 98,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Inspiration from: https://pandas.pydata.org/docs/user_guide/merging.html.\n",
    "\n",
    "class ImagesAndLabelsDataset(Dataset): \n",
    "    def __init__(self, csv_file, root_dir, transform=None):\n",
    "        self.annotations = pd.read_csv(csv_file)      # labels\n",
    "        self.root_dir = root_dir\n",
    "        self.transform = transform\n",
    "        self.scaler = sklearn.preprocessing.StandardScaler()\n",
    "        \n",
    "        # Fit the scaler to VMAF scores\n",
    "        self.scaler.fit(self.annotations.iloc[:, 0].values.reshape(-1, 1))\n",
    "        \n",
    "    def __len__(self):\n",
    "        return len(self.annotations)\n",
    "    \n",
    "    def __getimage__(self, index):\n",
    "        img_path = self.annotations.iloc[index, 1]\n",
    "        return img_path\n",
    "    \n",
    "    def __getitem__(self, index):\n",
    "        img_path = self.annotations.iloc[index, 1]\n",
    "        try:\n",
    "            image = mpimg.imread(img_path)\n",
    "        except Exception as e:\n",
    "            print(f\"Error reading image at index {index}: {e}\")\n",
    "            return None\n",
    "        \n",
    "        \n",
    "        y_label = self.annotations.iloc[index, 0]  # VMAF score\n",
    "        \n",
    "        # Normalize y_label using the previously created scaler\n",
    "        normalized_vmaf = self.scaler.transform([[y_label]])[0][0]\n",
    "        \n",
    "        normalized_vmaf = torch.tensor(normalized_vmaf, dtype=torch.float32)   # standardized VMAF (between -1 and 1)\n",
    "        norm = transforms.Compose([transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])   # normalization, from https://pytorch.org/hub/pytorch_vision_densenet/\n",
    "        \n",
    "        if self.transform:\n",
    "            image = self.transform(image)\n",
    "            image = norm(image)\n",
    "        \n",
    "            \n",
    "        return (image, normalized_vmaf, img_path)     # tuple of tensors and img_path\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(tensor([[[-2.1179, -2.1179, -2.1179,  ..., -2.1179, -2.1179, -2.1179],\n",
      "         [-2.1179, -2.1179, -2.1179,  ..., -2.1179, -2.1179, -2.1179],\n",
      "         [-2.1179, -2.1179, -2.1179,  ..., -2.1179, -2.1179, -2.1179],\n",
      "         ...,\n",
      "         [-2.1179, -2.1179, -2.1179,  ..., -2.1179, -2.1179, -2.1179],\n",
      "         [-2.1179, -2.1179, -2.1179,  ..., -2.1179, -2.1179, -2.1179],\n",
      "         [-2.1179, -2.1179, -2.1179,  ..., -2.1179, -2.1179, -2.1179]],\n",
      "\n",
      "        [[-2.0357, -2.0357, -2.0357,  ..., -2.0357, -2.0357, -2.0357],\n",
      "         [-2.0357, -2.0357, -2.0357,  ..., -2.0357, -2.0357, -2.0357],\n",
      "         [-2.0357, -2.0357, -2.0357,  ..., -2.0357, -2.0357, -2.0357],\n",
      "         ...,\n",
      "         [-2.0357, -2.0357, -2.0357,  ..., -2.0357, -2.0357, -2.0357],\n",
      "         [-2.0357, -2.0357, -2.0357,  ..., -2.0357, -2.0357, -2.0357],\n",
      "         [-2.0357, -2.0357, -2.0357,  ..., -2.0357, -2.0357, -2.0357]],\n",
      "\n",
      "        [[-1.8044, -1.8044, -1.8044,  ..., -1.8044, -1.8044, -1.8044],\n",
      "         [-1.8044, -1.8044, -1.8044,  ..., -1.8044, -1.8044, -1.8044],\n",
      "         [-1.8044, -1.8044, -1.8044,  ..., -1.8044, -1.8044, -1.8044],\n",
      "         ...,\n",
      "         [-1.8044, -1.8044, -1.8044,  ..., -1.8044, -1.8044, -1.8044],\n",
      "         [-1.8044, -1.8044, -1.8044,  ..., -1.8044, -1.8044, -1.8044],\n",
      "         [-1.8044, -1.8044, -1.8044,  ..., -1.8044, -1.8044, -1.8044]]]), tensor(1.0642), '/Users/kajsalidin/Desktop/nr-vqa-vmaf/images_train_comp_crop_ALL/train_1_compressed_23_a-hidden-life.mp4_frame0_left__start_0_224.png')\n"
     ]
    }
   ],
   "source": [
    "data = ImagesAndLabelsDataset(csv_file = \"vmaf_values_train_ALL.csv\", root_dir = \"images_train_comp_crop_ALL\", transform = transforms.ToTensor())\n",
    "print(data[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_set, val_set = torch.utils.data.random_split(data, [0.9, 0.1], generator=torch.Generator().manual_seed(3))    # 10% validation \n",
    "train_loader = DataLoader(dataset=train_set, shuffle=True, batch_size=100)   # Shuffle data.\n",
    "val_loader = DataLoader(dataset=val_set, shuffle=True, batch_size=100)\n",
    "\n",
    "# Data loader combines a dataset and a sampler, and provides an iterable over the given dataset, easy access to the samples."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "70071\n",
      "701\n"
     ]
    }
   ],
   "source": [
    "print(len(train_set))\n",
    "print(len(train_loader))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/Users/kajsalidin/Desktop/nr-vqa-vmaf/images_TEST_comp_crop_png_ALL/TEST_313_compressed_37_robocop-2.mp4_frame0_top_start_224_0.png\n"
     ]
    }
   ],
   "source": [
    "# TODO: standardize test data\n",
    "test_data = ImagesAndLabelsDataset(csv_file = \"vmaf_values_TEST_ALL.csv\", root_dir = \"images_TEST_comp_crop_png_ALL\", transform = transforms.ToTensor())\n",
    "print(test_data.__getimage__(10000)) \n",
    "# test_loader = DataLoader(dataset=test_data, shuffle=False, batch_size=100)   # kanske till och med ännu större batch_size för testset?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# New test, not used for study.\n",
    "# Separate DenseNet embeddings from dense layers in Classifier class below.\n",
    "# Inspired from https://www.kaggle.com/code/leighplt/densenet121-pytorch.\n",
    "class DenseNet(nn.Module):\n",
    "    def __init__(self, num_channels=3):\n",
    "        super().__init__()\n",
    "        preloaded = models.densenet121(pretrained=True)\n",
    "        self.features = preloaded.features\n",
    "        self.features.conv0 = nn.Conv2d(num_channels, 64, 7, 2, 3)   # in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', device=None, dtype=None\n",
    "        \n",
    "        for param in self.features.parameters():  # Freeze all parameters in self.features\n",
    "            param.requires_grad = False\n",
    "\n",
    "          \n",
    "    def forward(self, x):\n",
    "        features = self.features(x)\n",
    "        return features\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Classifier(nn.Module):\n",
    "  def __init__(self):\n",
    "      super().__init__()\n",
    "      self.dense = nn.Linear(50176, 1024, bias=True)\n",
    "      self.dense2 = nn.Linear(1024, 1024, bias=True)\n",
    "      self.dense3 = nn.Linear(1024, 1024, bias=True)\n",
    "      self.dense4 = nn.Linear(1024, 1024, bias=True)\n",
    "      self.dense5 = nn.Linear(1024, 1024, bias=True)\n",
    "      self.classifier = nn.Linear(1024, 1, bias=True)\n",
    "\n",
    "  def forward(self, x):\n",
    "      out = torch.flatten(x, 1)       \n",
    "      out = self.dense(out)\n",
    "      out = F.leaky_relu(out)   \n",
    "      out = self.dense2(out)\n",
    "      out = F.leaky_relu(out)\n",
    "      out = self.dense3(out)\n",
    "      out = F.leaky_relu(out)\n",
    "      out = self.dense4(out)\n",
    "      out = F.leaky_relu(out)\n",
    "      out = self.dense5(out)\n",
    "      out = F.leaky_relu(out)\n",
    "      out = self.classifier(out)\n",
    "      return out.squeeze(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# New test, not used for study.\n",
    "# Inspiration from: https://pandas.pydata.org/docs/user_guide/merging.html.\n",
    "\n",
    "class ImagesAndLabelsDataset(Dataset):\n",
    "    def __init__(self, csv_file, root_dir, scaler, transform=None):\n",
    "        self.annotations = pd.read_csv(csv_file)      # labels\n",
    "        self.root_dir = root_dir\n",
    "        self.transform = transform\n",
    "        self.scaler = scaler\n",
    "\n",
    "        # Fit the scaler to VMAF scores.\n",
    "        self.scaler.fit(self.annotations.iloc[:, 0].values.reshape(-1, 1))\n",
    "\n",
    "    def __len__(self):\n",
    "        return len(self.annotations)\n",
    "\n",
    "    def __getimage__(self, index):\n",
    "        img_path = self.annotations.iloc[index, 1]\n",
    "        return img_path\n",
    "\n",
    "    def __getitem__(self, index):\n",
    "        img_path = self.annotations.iloc[index, 1]\n",
    "        try:\n",
    "            image_output = mpimg.imread(img_path)    # pixel intensities between 0-1.\n",
    "        except Exception as e:\n",
    "            print(f\"Error reading image at index {index}: {e}\")\n",
    "            return None\n",
    "\n",
    "\n",
    "        y_label = self.annotations.iloc[index, 0]  # VMAF score.\n",
    "\n",
    "        # Normalize y_label using the previously created scaler.\n",
    "        normalized_vmaf = self.scaler.transform([[y_label]])[0][0]\n",
    "\n",
    "        normalized_vmaf = torch.Tensor([normalized_vmaf])   \n",
    "        norm = transforms.Compose([transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])   # ImageNet normalization, from https://pytorch.org/hub/pytorch_vision_densenet/.\n",
    "\n",
    "        if self.transform:\n",
    "            image_output = self.transform(image_output)\n",
    "            image_output = norm(image_output)      \n",
    "            image_output = image_output.unsqueeze(0)\n",
    "            model = DenseNet(num_channels=3)\n",
    "            densenet_output = model(image_output)  # DenseNet embeddings.\n",
    "\n",
    "\n",
    "        return (densenet_output, normalized_vmaf, img_path)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(tensor([-0.0214], grad_fn=<SqueezeBackward1>), tensor([1.0642]), '/Users/kajsalidin/Desktop/nr-vqa-vmaf/images_train_comp_crop_ALL/train_1_compressed_23_a-hidden-life.mp4_frame0_left_top_start_0_0.png')\n"
     ]
    }
   ],
   "source": [
    "data = ImagesAndLabelsDataset(csv_file = \"vmaf_values_train_ALL.csv\", root_dir = \"images_train_comp_crop_ALL\", scaler=sklearn.preprocessing.StandardScaler(), transform = transforms.ToTensor())\n",
    "print(data[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_set, val_set = torch.utils.data.random_split(data, [0.9, 0.1], generator=torch.Generator().manual_seed(3))    # 10% validation\n",
    "train_loader = DataLoader(dataset=train_set, shuffle=True, batch_size=10) \n",
    "val_loader = DataLoader(dataset=val_set, shuffle=True, batch_size=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/Users/kajsalidin/Desktop/nr-vqa-vmaf/images_TEST_comp_crop_png_ALL/TEST_1_compressed_23_agent-cody-banks.mp4_frame0__start_1344_448.png\n"
     ]
    }
   ],
   "source": [
    "test_data = ImagesAndLabelsDataset(csv_file = \"vmaf_values_TEST_ALL.csv\", root_dir = \"images_TEST_comp_crop_png_ALL\", scaler= sklearn.preprocessing.StandardScaler(), transform = transforms.ToTensor())\n",
    "print(test_data.__getimage__(10))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "features.conv0.weight \t torch.Size([64, 3, 7, 7])\n",
      "features.conv0.bias \t torch.Size([64])\n",
      "features.norm0.weight \t torch.Size([64])\n",
      "features.norm0.bias \t torch.Size([64])\n",
      "features.norm0.running_mean \t torch.Size([64])\n",
      "features.norm0.running_var \t torch.Size([64])\n",
      "features.norm0.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock1.denselayer1.norm1.weight \t torch.Size([64])\n",
      "features.denseblock1.denselayer1.norm1.bias \t torch.Size([64])\n",
      "features.denseblock1.denselayer1.norm1.running_mean \t torch.Size([64])\n",
      "features.denseblock1.denselayer1.norm1.running_var \t torch.Size([64])\n",
      "features.denseblock1.denselayer1.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock1.denselayer1.conv1.weight \t torch.Size([128, 64, 1, 1])\n",
      "features.denseblock1.denselayer1.norm2.weight \t torch.Size([128])\n",
      "features.denseblock1.denselayer1.norm2.bias \t torch.Size([128])\n",
      "features.denseblock1.denselayer1.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock1.denselayer1.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock1.denselayer1.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock1.denselayer1.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock1.denselayer2.norm1.weight \t torch.Size([96])\n",
      "features.denseblock1.denselayer2.norm1.bias \t torch.Size([96])\n",
      "features.denseblock1.denselayer2.norm1.running_mean \t torch.Size([96])\n",
      "features.denseblock1.denselayer2.norm1.running_var \t torch.Size([96])\n",
      "features.denseblock1.denselayer2.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock1.denselayer2.conv1.weight \t torch.Size([128, 96, 1, 1])\n",
      "features.denseblock1.denselayer2.norm2.weight \t torch.Size([128])\n",
      "features.denseblock1.denselayer2.norm2.bias \t torch.Size([128])\n",
      "features.denseblock1.denselayer2.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock1.denselayer2.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock1.denselayer2.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock1.denselayer2.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock1.denselayer3.norm1.weight \t torch.Size([128])\n",
      "features.denseblock1.denselayer3.norm1.bias \t torch.Size([128])\n",
      "features.denseblock1.denselayer3.norm1.running_mean \t torch.Size([128])\n",
      "features.denseblock1.denselayer3.norm1.running_var \t torch.Size([128])\n",
      "features.denseblock1.denselayer3.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock1.denselayer3.conv1.weight \t torch.Size([128, 128, 1, 1])\n",
      "features.denseblock1.denselayer3.norm2.weight \t torch.Size([128])\n",
      "features.denseblock1.denselayer3.norm2.bias \t torch.Size([128])\n",
      "features.denseblock1.denselayer3.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock1.denselayer3.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock1.denselayer3.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock1.denselayer3.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock1.denselayer4.norm1.weight \t torch.Size([160])\n",
      "features.denseblock1.denselayer4.norm1.bias \t torch.Size([160])\n",
      "features.denseblock1.denselayer4.norm1.running_mean \t torch.Size([160])\n",
      "features.denseblock1.denselayer4.norm1.running_var \t torch.Size([160])\n",
      "features.denseblock1.denselayer4.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock1.denselayer4.conv1.weight \t torch.Size([128, 160, 1, 1])\n",
      "features.denseblock1.denselayer4.norm2.weight \t torch.Size([128])\n",
      "features.denseblock1.denselayer4.norm2.bias \t torch.Size([128])\n",
      "features.denseblock1.denselayer4.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock1.denselayer4.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock1.denselayer4.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock1.denselayer4.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock1.denselayer5.norm1.weight \t torch.Size([192])\n",
      "features.denseblock1.denselayer5.norm1.bias \t torch.Size([192])\n",
      "features.denseblock1.denselayer5.norm1.running_mean \t torch.Size([192])\n",
      "features.denseblock1.denselayer5.norm1.running_var \t torch.Size([192])\n",
      "features.denseblock1.denselayer5.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock1.denselayer5.conv1.weight \t torch.Size([128, 192, 1, 1])\n",
      "features.denseblock1.denselayer5.norm2.weight \t torch.Size([128])\n",
      "features.denseblock1.denselayer5.norm2.bias \t torch.Size([128])\n",
      "features.denseblock1.denselayer5.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock1.denselayer5.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock1.denselayer5.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock1.denselayer5.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock1.denselayer6.norm1.weight \t torch.Size([224])\n",
      "features.denseblock1.denselayer6.norm1.bias \t torch.Size([224])\n",
      "features.denseblock1.denselayer6.norm1.running_mean \t torch.Size([224])\n",
      "features.denseblock1.denselayer6.norm1.running_var \t torch.Size([224])\n",
      "features.denseblock1.denselayer6.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock1.denselayer6.conv1.weight \t torch.Size([128, 224, 1, 1])\n",
      "features.denseblock1.denselayer6.norm2.weight \t torch.Size([128])\n",
      "features.denseblock1.denselayer6.norm2.bias \t torch.Size([128])\n",
      "features.denseblock1.denselayer6.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock1.denselayer6.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock1.denselayer6.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock1.denselayer6.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.transition1.norm.weight \t torch.Size([256])\n",
      "features.transition1.norm.bias \t torch.Size([256])\n",
      "features.transition1.norm.running_mean \t torch.Size([256])\n",
      "features.transition1.norm.running_var \t torch.Size([256])\n",
      "features.transition1.norm.num_batches_tracked \t torch.Size([])\n",
      "features.transition1.conv.weight \t torch.Size([128, 256, 1, 1])\n",
      "features.denseblock2.denselayer1.norm1.weight \t torch.Size([128])\n",
      "features.denseblock2.denselayer1.norm1.bias \t torch.Size([128])\n",
      "features.denseblock2.denselayer1.norm1.running_mean \t torch.Size([128])\n",
      "features.denseblock2.denselayer1.norm1.running_var \t torch.Size([128])\n",
      "features.denseblock2.denselayer1.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer1.conv1.weight \t torch.Size([128, 128, 1, 1])\n",
      "features.denseblock2.denselayer1.norm2.weight \t torch.Size([128])\n",
      "features.denseblock2.denselayer1.norm2.bias \t torch.Size([128])\n",
      "features.denseblock2.denselayer1.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock2.denselayer1.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock2.denselayer1.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer1.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock2.denselayer2.norm1.weight \t torch.Size([160])\n",
      "features.denseblock2.denselayer2.norm1.bias \t torch.Size([160])\n",
      "features.denseblock2.denselayer2.norm1.running_mean \t torch.Size([160])\n",
      "features.denseblock2.denselayer2.norm1.running_var \t torch.Size([160])\n",
      "features.denseblock2.denselayer2.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer2.conv1.weight \t torch.Size([128, 160, 1, 1])\n",
      "features.denseblock2.denselayer2.norm2.weight \t torch.Size([128])\n",
      "features.denseblock2.denselayer2.norm2.bias \t torch.Size([128])\n",
      "features.denseblock2.denselayer2.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock2.denselayer2.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock2.denselayer2.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer2.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock2.denselayer3.norm1.weight \t torch.Size([192])\n",
      "features.denseblock2.denselayer3.norm1.bias \t torch.Size([192])\n",
      "features.denseblock2.denselayer3.norm1.running_mean \t torch.Size([192])\n",
      "features.denseblock2.denselayer3.norm1.running_var \t torch.Size([192])\n",
      "features.denseblock2.denselayer3.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer3.conv1.weight \t torch.Size([128, 192, 1, 1])\n",
      "features.denseblock2.denselayer3.norm2.weight \t torch.Size([128])\n",
      "features.denseblock2.denselayer3.norm2.bias \t torch.Size([128])\n",
      "features.denseblock2.denselayer3.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock2.denselayer3.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock2.denselayer3.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer3.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock2.denselayer4.norm1.weight \t torch.Size([224])\n",
      "features.denseblock2.denselayer4.norm1.bias \t torch.Size([224])\n",
      "features.denseblock2.denselayer4.norm1.running_mean \t torch.Size([224])\n",
      "features.denseblock2.denselayer4.norm1.running_var \t torch.Size([224])\n",
      "features.denseblock2.denselayer4.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer4.conv1.weight \t torch.Size([128, 224, 1, 1])\n",
      "features.denseblock2.denselayer4.norm2.weight \t torch.Size([128])\n",
      "features.denseblock2.denselayer4.norm2.bias \t torch.Size([128])\n",
      "features.denseblock2.denselayer4.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock2.denselayer4.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock2.denselayer4.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer4.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock2.denselayer5.norm1.weight \t torch.Size([256])\n",
      "features.denseblock2.denselayer5.norm1.bias \t torch.Size([256])\n",
      "features.denseblock2.denselayer5.norm1.running_mean \t torch.Size([256])\n",
      "features.denseblock2.denselayer5.norm1.running_var \t torch.Size([256])\n",
      "features.denseblock2.denselayer5.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer5.conv1.weight \t torch.Size([128, 256, 1, 1])\n",
      "features.denseblock2.denselayer5.norm2.weight \t torch.Size([128])\n",
      "features.denseblock2.denselayer5.norm2.bias \t torch.Size([128])\n",
      "features.denseblock2.denselayer5.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock2.denselayer5.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock2.denselayer5.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer5.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock2.denselayer6.norm1.weight \t torch.Size([288])\n",
      "features.denseblock2.denselayer6.norm1.bias \t torch.Size([288])\n",
      "features.denseblock2.denselayer6.norm1.running_mean \t torch.Size([288])\n",
      "features.denseblock2.denselayer6.norm1.running_var \t torch.Size([288])\n",
      "features.denseblock2.denselayer6.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer6.conv1.weight \t torch.Size([128, 288, 1, 1])\n",
      "features.denseblock2.denselayer6.norm2.weight \t torch.Size([128])\n",
      "features.denseblock2.denselayer6.norm2.bias \t torch.Size([128])\n",
      "features.denseblock2.denselayer6.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock2.denselayer6.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock2.denselayer6.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer6.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock2.denselayer7.norm1.weight \t torch.Size([320])\n",
      "features.denseblock2.denselayer7.norm1.bias \t torch.Size([320])\n",
      "features.denseblock2.denselayer7.norm1.running_mean \t torch.Size([320])\n",
      "features.denseblock2.denselayer7.norm1.running_var \t torch.Size([320])\n",
      "features.denseblock2.denselayer7.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer7.conv1.weight \t torch.Size([128, 320, 1, 1])\n",
      "features.denseblock2.denselayer7.norm2.weight \t torch.Size([128])\n",
      "features.denseblock2.denselayer7.norm2.bias \t torch.Size([128])\n",
      "features.denseblock2.denselayer7.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock2.denselayer7.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock2.denselayer7.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer7.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock2.denselayer8.norm1.weight \t torch.Size([352])\n",
      "features.denseblock2.denselayer8.norm1.bias \t torch.Size([352])\n",
      "features.denseblock2.denselayer8.norm1.running_mean \t torch.Size([352])\n",
      "features.denseblock2.denselayer8.norm1.running_var \t torch.Size([352])\n",
      "features.denseblock2.denselayer8.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer8.conv1.weight \t torch.Size([128, 352, 1, 1])\n",
      "features.denseblock2.denselayer8.norm2.weight \t torch.Size([128])\n",
      "features.denseblock2.denselayer8.norm2.bias \t torch.Size([128])\n",
      "features.denseblock2.denselayer8.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock2.denselayer8.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock2.denselayer8.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer8.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock2.denselayer9.norm1.weight \t torch.Size([384])\n",
      "features.denseblock2.denselayer9.norm1.bias \t torch.Size([384])\n",
      "features.denseblock2.denselayer9.norm1.running_mean \t torch.Size([384])\n",
      "features.denseblock2.denselayer9.norm1.running_var \t torch.Size([384])\n",
      "features.denseblock2.denselayer9.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer9.conv1.weight \t torch.Size([128, 384, 1, 1])\n",
      "features.denseblock2.denselayer9.norm2.weight \t torch.Size([128])\n",
      "features.denseblock2.denselayer9.norm2.bias \t torch.Size([128])\n",
      "features.denseblock2.denselayer9.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock2.denselayer9.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock2.denselayer9.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer9.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock2.denselayer10.norm1.weight \t torch.Size([416])\n",
      "features.denseblock2.denselayer10.norm1.bias \t torch.Size([416])\n",
      "features.denseblock2.denselayer10.norm1.running_mean \t torch.Size([416])\n",
      "features.denseblock2.denselayer10.norm1.running_var \t torch.Size([416])\n",
      "features.denseblock2.denselayer10.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer10.conv1.weight \t torch.Size([128, 416, 1, 1])\n",
      "features.denseblock2.denselayer10.norm2.weight \t torch.Size([128])\n",
      "features.denseblock2.denselayer10.norm2.bias \t torch.Size([128])\n",
      "features.denseblock2.denselayer10.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock2.denselayer10.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock2.denselayer10.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer10.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock2.denselayer11.norm1.weight \t torch.Size([448])\n",
      "features.denseblock2.denselayer11.norm1.bias \t torch.Size([448])\n",
      "features.denseblock2.denselayer11.norm1.running_mean \t torch.Size([448])\n",
      "features.denseblock2.denselayer11.norm1.running_var \t torch.Size([448])\n",
      "features.denseblock2.denselayer11.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer11.conv1.weight \t torch.Size([128, 448, 1, 1])\n",
      "features.denseblock2.denselayer11.norm2.weight \t torch.Size([128])\n",
      "features.denseblock2.denselayer11.norm2.bias \t torch.Size([128])\n",
      "features.denseblock2.denselayer11.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock2.denselayer11.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock2.denselayer11.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer11.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock2.denselayer12.norm1.weight \t torch.Size([480])\n",
      "features.denseblock2.denselayer12.norm1.bias \t torch.Size([480])\n",
      "features.denseblock2.denselayer12.norm1.running_mean \t torch.Size([480])\n",
      "features.denseblock2.denselayer12.norm1.running_var \t torch.Size([480])\n",
      "features.denseblock2.denselayer12.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer12.conv1.weight \t torch.Size([128, 480, 1, 1])\n",
      "features.denseblock2.denselayer12.norm2.weight \t torch.Size([128])\n",
      "features.denseblock2.denselayer12.norm2.bias \t torch.Size([128])\n",
      "features.denseblock2.denselayer12.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock2.denselayer12.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock2.denselayer12.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock2.denselayer12.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.transition2.norm.weight \t torch.Size([512])\n",
      "features.transition2.norm.bias \t torch.Size([512])\n",
      "features.transition2.norm.running_mean \t torch.Size([512])\n",
      "features.transition2.norm.running_var \t torch.Size([512])\n",
      "features.transition2.norm.num_batches_tracked \t torch.Size([])\n",
      "features.transition2.conv.weight \t torch.Size([256, 512, 1, 1])\n",
      "features.denseblock3.denselayer1.norm1.weight \t torch.Size([256])\n",
      "features.denseblock3.denselayer1.norm1.bias \t torch.Size([256])\n",
      "features.denseblock3.denselayer1.norm1.running_mean \t torch.Size([256])\n",
      "features.denseblock3.denselayer1.norm1.running_var \t torch.Size([256])\n",
      "features.denseblock3.denselayer1.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer1.conv1.weight \t torch.Size([128, 256, 1, 1])\n",
      "features.denseblock3.denselayer1.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer1.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer1.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer1.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer1.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer1.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer2.norm1.weight \t torch.Size([288])\n",
      "features.denseblock3.denselayer2.norm1.bias \t torch.Size([288])\n",
      "features.denseblock3.denselayer2.norm1.running_mean \t torch.Size([288])\n",
      "features.denseblock3.denselayer2.norm1.running_var \t torch.Size([288])\n",
      "features.denseblock3.denselayer2.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer2.conv1.weight \t torch.Size([128, 288, 1, 1])\n",
      "features.denseblock3.denselayer2.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer2.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer2.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer2.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer2.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer2.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer3.norm1.weight \t torch.Size([320])\n",
      "features.denseblock3.denselayer3.norm1.bias \t torch.Size([320])\n",
      "features.denseblock3.denselayer3.norm1.running_mean \t torch.Size([320])\n",
      "features.denseblock3.denselayer3.norm1.running_var \t torch.Size([320])\n",
      "features.denseblock3.denselayer3.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer3.conv1.weight \t torch.Size([128, 320, 1, 1])\n",
      "features.denseblock3.denselayer3.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer3.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer3.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer3.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer3.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer3.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer4.norm1.weight \t torch.Size([352])\n",
      "features.denseblock3.denselayer4.norm1.bias \t torch.Size([352])\n",
      "features.denseblock3.denselayer4.norm1.running_mean \t torch.Size([352])\n",
      "features.denseblock3.denselayer4.norm1.running_var \t torch.Size([352])\n",
      "features.denseblock3.denselayer4.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer4.conv1.weight \t torch.Size([128, 352, 1, 1])\n",
      "features.denseblock3.denselayer4.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer4.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer4.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer4.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer4.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer4.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer5.norm1.weight \t torch.Size([384])\n",
      "features.denseblock3.denselayer5.norm1.bias \t torch.Size([384])\n",
      "features.denseblock3.denselayer5.norm1.running_mean \t torch.Size([384])\n",
      "features.denseblock3.denselayer5.norm1.running_var \t torch.Size([384])\n",
      "features.denseblock3.denselayer5.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer5.conv1.weight \t torch.Size([128, 384, 1, 1])\n",
      "features.denseblock3.denselayer5.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer5.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer5.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer5.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer5.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer5.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer6.norm1.weight \t torch.Size([416])\n",
      "features.denseblock3.denselayer6.norm1.bias \t torch.Size([416])\n",
      "features.denseblock3.denselayer6.norm1.running_mean \t torch.Size([416])\n",
      "features.denseblock3.denselayer6.norm1.running_var \t torch.Size([416])\n",
      "features.denseblock3.denselayer6.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer6.conv1.weight \t torch.Size([128, 416, 1, 1])\n",
      "features.denseblock3.denselayer6.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer6.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer6.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer6.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer6.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer6.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer7.norm1.weight \t torch.Size([448])\n",
      "features.denseblock3.denselayer7.norm1.bias \t torch.Size([448])\n",
      "features.denseblock3.denselayer7.norm1.running_mean \t torch.Size([448])\n",
      "features.denseblock3.denselayer7.norm1.running_var \t torch.Size([448])\n",
      "features.denseblock3.denselayer7.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer7.conv1.weight \t torch.Size([128, 448, 1, 1])\n",
      "features.denseblock3.denselayer7.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer7.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer7.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer7.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer7.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer7.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer8.norm1.weight \t torch.Size([480])\n",
      "features.denseblock3.denselayer8.norm1.bias \t torch.Size([480])\n",
      "features.denseblock3.denselayer8.norm1.running_mean \t torch.Size([480])\n",
      "features.denseblock3.denselayer8.norm1.running_var \t torch.Size([480])\n",
      "features.denseblock3.denselayer8.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer8.conv1.weight \t torch.Size([128, 480, 1, 1])\n",
      "features.denseblock3.denselayer8.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer8.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer8.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer8.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer8.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer8.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer9.norm1.weight \t torch.Size([512])\n",
      "features.denseblock3.denselayer9.norm1.bias \t torch.Size([512])\n",
      "features.denseblock3.denselayer9.norm1.running_mean \t torch.Size([512])\n",
      "features.denseblock3.denselayer9.norm1.running_var \t torch.Size([512])\n",
      "features.denseblock3.denselayer9.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer9.conv1.weight \t torch.Size([128, 512, 1, 1])\n",
      "features.denseblock3.denselayer9.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer9.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer9.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer9.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer9.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer9.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer10.norm1.weight \t torch.Size([544])\n",
      "features.denseblock3.denselayer10.norm1.bias \t torch.Size([544])\n",
      "features.denseblock3.denselayer10.norm1.running_mean \t torch.Size([544])\n",
      "features.denseblock3.denselayer10.norm1.running_var \t torch.Size([544])\n",
      "features.denseblock3.denselayer10.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer10.conv1.weight \t torch.Size([128, 544, 1, 1])\n",
      "features.denseblock3.denselayer10.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer10.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer10.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer10.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer10.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer10.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer11.norm1.weight \t torch.Size([576])\n",
      "features.denseblock3.denselayer11.norm1.bias \t torch.Size([576])\n",
      "features.denseblock3.denselayer11.norm1.running_mean \t torch.Size([576])\n",
      "features.denseblock3.denselayer11.norm1.running_var \t torch.Size([576])\n",
      "features.denseblock3.denselayer11.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer11.conv1.weight \t torch.Size([128, 576, 1, 1])\n",
      "features.denseblock3.denselayer11.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer11.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer11.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer11.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer11.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer11.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer12.norm1.weight \t torch.Size([608])\n",
      "features.denseblock3.denselayer12.norm1.bias \t torch.Size([608])\n",
      "features.denseblock3.denselayer12.norm1.running_mean \t torch.Size([608])\n",
      "features.denseblock3.denselayer12.norm1.running_var \t torch.Size([608])\n",
      "features.denseblock3.denselayer12.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer12.conv1.weight \t torch.Size([128, 608, 1, 1])\n",
      "features.denseblock3.denselayer12.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer12.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer12.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer12.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer12.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer12.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer13.norm1.weight \t torch.Size([640])\n",
      "features.denseblock3.denselayer13.norm1.bias \t torch.Size([640])\n",
      "features.denseblock3.denselayer13.norm1.running_mean \t torch.Size([640])\n",
      "features.denseblock3.denselayer13.norm1.running_var \t torch.Size([640])\n",
      "features.denseblock3.denselayer13.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer13.conv1.weight \t torch.Size([128, 640, 1, 1])\n",
      "features.denseblock3.denselayer13.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer13.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer13.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer13.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer13.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer13.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer14.norm1.weight \t torch.Size([672])\n",
      "features.denseblock3.denselayer14.norm1.bias \t torch.Size([672])\n",
      "features.denseblock3.denselayer14.norm1.running_mean \t torch.Size([672])\n",
      "features.denseblock3.denselayer14.norm1.running_var \t torch.Size([672])\n",
      "features.denseblock3.denselayer14.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer14.conv1.weight \t torch.Size([128, 672, 1, 1])\n",
      "features.denseblock3.denselayer14.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer14.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer14.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer14.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer14.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer14.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer15.norm1.weight \t torch.Size([704])\n",
      "features.denseblock3.denselayer15.norm1.bias \t torch.Size([704])\n",
      "features.denseblock3.denselayer15.norm1.running_mean \t torch.Size([704])\n",
      "features.denseblock3.denselayer15.norm1.running_var \t torch.Size([704])\n",
      "features.denseblock3.denselayer15.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer15.conv1.weight \t torch.Size([128, 704, 1, 1])\n",
      "features.denseblock3.denselayer15.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer15.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer15.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer15.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer15.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer15.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer16.norm1.weight \t torch.Size([736])\n",
      "features.denseblock3.denselayer16.norm1.bias \t torch.Size([736])\n",
      "features.denseblock3.denselayer16.norm1.running_mean \t torch.Size([736])\n",
      "features.denseblock3.denselayer16.norm1.running_var \t torch.Size([736])\n",
      "features.denseblock3.denselayer16.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer16.conv1.weight \t torch.Size([128, 736, 1, 1])\n",
      "features.denseblock3.denselayer16.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer16.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer16.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer16.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer16.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer16.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer17.norm1.weight \t torch.Size([768])\n",
      "features.denseblock3.denselayer17.norm1.bias \t torch.Size([768])\n",
      "features.denseblock3.denselayer17.norm1.running_mean \t torch.Size([768])\n",
      "features.denseblock3.denselayer17.norm1.running_var \t torch.Size([768])\n",
      "features.denseblock3.denselayer17.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer17.conv1.weight \t torch.Size([128, 768, 1, 1])\n",
      "features.denseblock3.denselayer17.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer17.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer17.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer17.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer17.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer17.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer18.norm1.weight \t torch.Size([800])\n",
      "features.denseblock3.denselayer18.norm1.bias \t torch.Size([800])\n",
      "features.denseblock3.denselayer18.norm1.running_mean \t torch.Size([800])\n",
      "features.denseblock3.denselayer18.norm1.running_var \t torch.Size([800])\n",
      "features.denseblock3.denselayer18.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer18.conv1.weight \t torch.Size([128, 800, 1, 1])\n",
      "features.denseblock3.denselayer18.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer18.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer18.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer18.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer18.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer18.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer19.norm1.weight \t torch.Size([832])\n",
      "features.denseblock3.denselayer19.norm1.bias \t torch.Size([832])\n",
      "features.denseblock3.denselayer19.norm1.running_mean \t torch.Size([832])\n",
      "features.denseblock3.denselayer19.norm1.running_var \t torch.Size([832])\n",
      "features.denseblock3.denselayer19.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer19.conv1.weight \t torch.Size([128, 832, 1, 1])\n",
      "features.denseblock3.denselayer19.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer19.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer19.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer19.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer19.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer19.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer20.norm1.weight \t torch.Size([864])\n",
      "features.denseblock3.denselayer20.norm1.bias \t torch.Size([864])\n",
      "features.denseblock3.denselayer20.norm1.running_mean \t torch.Size([864])\n",
      "features.denseblock3.denselayer20.norm1.running_var \t torch.Size([864])\n",
      "features.denseblock3.denselayer20.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer20.conv1.weight \t torch.Size([128, 864, 1, 1])\n",
      "features.denseblock3.denselayer20.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer20.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer20.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer20.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer20.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer20.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer21.norm1.weight \t torch.Size([896])\n",
      "features.denseblock3.denselayer21.norm1.bias \t torch.Size([896])\n",
      "features.denseblock3.denselayer21.norm1.running_mean \t torch.Size([896])\n",
      "features.denseblock3.denselayer21.norm1.running_var \t torch.Size([896])\n",
      "features.denseblock3.denselayer21.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer21.conv1.weight \t torch.Size([128, 896, 1, 1])\n",
      "features.denseblock3.denselayer21.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer21.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer21.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer21.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer21.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer21.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer22.norm1.weight \t torch.Size([928])\n",
      "features.denseblock3.denselayer22.norm1.bias \t torch.Size([928])\n",
      "features.denseblock3.denselayer22.norm1.running_mean \t torch.Size([928])\n",
      "features.denseblock3.denselayer22.norm1.running_var \t torch.Size([928])\n",
      "features.denseblock3.denselayer22.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer22.conv1.weight \t torch.Size([128, 928, 1, 1])\n",
      "features.denseblock3.denselayer22.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer22.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer22.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer22.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer22.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer22.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer23.norm1.weight \t torch.Size([960])\n",
      "features.denseblock3.denselayer23.norm1.bias \t torch.Size([960])\n",
      "features.denseblock3.denselayer23.norm1.running_mean \t torch.Size([960])\n",
      "features.denseblock3.denselayer23.norm1.running_var \t torch.Size([960])\n",
      "features.denseblock3.denselayer23.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer23.conv1.weight \t torch.Size([128, 960, 1, 1])\n",
      "features.denseblock3.denselayer23.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer23.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer23.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer23.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer23.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer23.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock3.denselayer24.norm1.weight \t torch.Size([992])\n",
      "features.denseblock3.denselayer24.norm1.bias \t torch.Size([992])\n",
      "features.denseblock3.denselayer24.norm1.running_mean \t torch.Size([992])\n",
      "features.denseblock3.denselayer24.norm1.running_var \t torch.Size([992])\n",
      "features.denseblock3.denselayer24.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer24.conv1.weight \t torch.Size([128, 992, 1, 1])\n",
      "features.denseblock3.denselayer24.norm2.weight \t torch.Size([128])\n",
      "features.denseblock3.denselayer24.norm2.bias \t torch.Size([128])\n",
      "features.denseblock3.denselayer24.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock3.denselayer24.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock3.denselayer24.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock3.denselayer24.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.transition3.norm.weight \t torch.Size([1024])\n",
      "features.transition3.norm.bias \t torch.Size([1024])\n",
      "features.transition3.norm.running_mean \t torch.Size([1024])\n",
      "features.transition3.norm.running_var \t torch.Size([1024])\n",
      "features.transition3.norm.num_batches_tracked \t torch.Size([])\n",
      "features.transition3.conv.weight \t torch.Size([512, 1024, 1, 1])\n",
      "features.denseblock4.denselayer1.norm1.weight \t torch.Size([512])\n",
      "features.denseblock4.denselayer1.norm1.bias \t torch.Size([512])\n",
      "features.denseblock4.denselayer1.norm1.running_mean \t torch.Size([512])\n",
      "features.denseblock4.denselayer1.norm1.running_var \t torch.Size([512])\n",
      "features.denseblock4.denselayer1.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer1.conv1.weight \t torch.Size([128, 512, 1, 1])\n",
      "features.denseblock4.denselayer1.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer1.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer1.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer1.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer1.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer1.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer2.norm1.weight \t torch.Size([544])\n",
      "features.denseblock4.denselayer2.norm1.bias \t torch.Size([544])\n",
      "features.denseblock4.denselayer2.norm1.running_mean \t torch.Size([544])\n",
      "features.denseblock4.denselayer2.norm1.running_var \t torch.Size([544])\n",
      "features.denseblock4.denselayer2.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer2.conv1.weight \t torch.Size([128, 544, 1, 1])\n",
      "features.denseblock4.denselayer2.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer2.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer2.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer2.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer2.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer2.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer3.norm1.weight \t torch.Size([576])\n",
      "features.denseblock4.denselayer3.norm1.bias \t torch.Size([576])\n",
      "features.denseblock4.denselayer3.norm1.running_mean \t torch.Size([576])\n",
      "features.denseblock4.denselayer3.norm1.running_var \t torch.Size([576])\n",
      "features.denseblock4.denselayer3.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer3.conv1.weight \t torch.Size([128, 576, 1, 1])\n",
      "features.denseblock4.denselayer3.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer3.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer3.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer3.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer3.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer3.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer4.norm1.weight \t torch.Size([608])\n",
      "features.denseblock4.denselayer4.norm1.bias \t torch.Size([608])\n",
      "features.denseblock4.denselayer4.norm1.running_mean \t torch.Size([608])\n",
      "features.denseblock4.denselayer4.norm1.running_var \t torch.Size([608])\n",
      "features.denseblock4.denselayer4.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer4.conv1.weight \t torch.Size([128, 608, 1, 1])\n",
      "features.denseblock4.denselayer4.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer4.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer4.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer4.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer4.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer4.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer5.norm1.weight \t torch.Size([640])\n",
      "features.denseblock4.denselayer5.norm1.bias \t torch.Size([640])\n",
      "features.denseblock4.denselayer5.norm1.running_mean \t torch.Size([640])\n",
      "features.denseblock4.denselayer5.norm1.running_var \t torch.Size([640])\n",
      "features.denseblock4.denselayer5.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer5.conv1.weight \t torch.Size([128, 640, 1, 1])\n",
      "features.denseblock4.denselayer5.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer5.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer5.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer5.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer5.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer5.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer6.norm1.weight \t torch.Size([672])\n",
      "features.denseblock4.denselayer6.norm1.bias \t torch.Size([672])\n",
      "features.denseblock4.denselayer6.norm1.running_mean \t torch.Size([672])\n",
      "features.denseblock4.denselayer6.norm1.running_var \t torch.Size([672])\n",
      "features.denseblock4.denselayer6.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer6.conv1.weight \t torch.Size([128, 672, 1, 1])\n",
      "features.denseblock4.denselayer6.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer6.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer6.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer6.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer6.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer6.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer7.norm1.weight \t torch.Size([704])\n",
      "features.denseblock4.denselayer7.norm1.bias \t torch.Size([704])\n",
      "features.denseblock4.denselayer7.norm1.running_mean \t torch.Size([704])\n",
      "features.denseblock4.denselayer7.norm1.running_var \t torch.Size([704])\n",
      "features.denseblock4.denselayer7.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer7.conv1.weight \t torch.Size([128, 704, 1, 1])\n",
      "features.denseblock4.denselayer7.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer7.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer7.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer7.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer7.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer7.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer8.norm1.weight \t torch.Size([736])\n",
      "features.denseblock4.denselayer8.norm1.bias \t torch.Size([736])\n",
      "features.denseblock4.denselayer8.norm1.running_mean \t torch.Size([736])\n",
      "features.denseblock4.denselayer8.norm1.running_var \t torch.Size([736])\n",
      "features.denseblock4.denselayer8.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer8.conv1.weight \t torch.Size([128, 736, 1, 1])\n",
      "features.denseblock4.denselayer8.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer8.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer8.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer8.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer8.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer8.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer9.norm1.weight \t torch.Size([768])\n",
      "features.denseblock4.denselayer9.norm1.bias \t torch.Size([768])\n",
      "features.denseblock4.denselayer9.norm1.running_mean \t torch.Size([768])\n",
      "features.denseblock4.denselayer9.norm1.running_var \t torch.Size([768])\n",
      "features.denseblock4.denselayer9.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer9.conv1.weight \t torch.Size([128, 768, 1, 1])\n",
      "features.denseblock4.denselayer9.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer9.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer9.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer9.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer9.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer9.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer10.norm1.weight \t torch.Size([800])\n",
      "features.denseblock4.denselayer10.norm1.bias \t torch.Size([800])\n",
      "features.denseblock4.denselayer10.norm1.running_mean \t torch.Size([800])\n",
      "features.denseblock4.denselayer10.norm1.running_var \t torch.Size([800])\n",
      "features.denseblock4.denselayer10.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer10.conv1.weight \t torch.Size([128, 800, 1, 1])\n",
      "features.denseblock4.denselayer10.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer10.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer10.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer10.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer10.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer10.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer11.norm1.weight \t torch.Size([832])\n",
      "features.denseblock4.denselayer11.norm1.bias \t torch.Size([832])\n",
      "features.denseblock4.denselayer11.norm1.running_mean \t torch.Size([832])\n",
      "features.denseblock4.denselayer11.norm1.running_var \t torch.Size([832])\n",
      "features.denseblock4.denselayer11.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer11.conv1.weight \t torch.Size([128, 832, 1, 1])\n",
      "features.denseblock4.denselayer11.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer11.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer11.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer11.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer11.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer11.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer12.norm1.weight \t torch.Size([864])\n",
      "features.denseblock4.denselayer12.norm1.bias \t torch.Size([864])\n",
      "features.denseblock4.denselayer12.norm1.running_mean \t torch.Size([864])\n",
      "features.denseblock4.denselayer12.norm1.running_var \t torch.Size([864])\n",
      "features.denseblock4.denselayer12.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer12.conv1.weight \t torch.Size([128, 864, 1, 1])\n",
      "features.denseblock4.denselayer12.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer12.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer12.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer12.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer12.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer12.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer13.norm1.weight \t torch.Size([896])\n",
      "features.denseblock4.denselayer13.norm1.bias \t torch.Size([896])\n",
      "features.denseblock4.denselayer13.norm1.running_mean \t torch.Size([896])\n",
      "features.denseblock4.denselayer13.norm1.running_var \t torch.Size([896])\n",
      "features.denseblock4.denselayer13.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer13.conv1.weight \t torch.Size([128, 896, 1, 1])\n",
      "features.denseblock4.denselayer13.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer13.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer13.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer13.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer13.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer13.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer14.norm1.weight \t torch.Size([928])\n",
      "features.denseblock4.denselayer14.norm1.bias \t torch.Size([928])\n",
      "features.denseblock4.denselayer14.norm1.running_mean \t torch.Size([928])\n",
      "features.denseblock4.denselayer14.norm1.running_var \t torch.Size([928])\n",
      "features.denseblock4.denselayer14.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer14.conv1.weight \t torch.Size([128, 928, 1, 1])\n",
      "features.denseblock4.denselayer14.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer14.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer14.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer14.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer14.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer14.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer15.norm1.weight \t torch.Size([960])\n",
      "features.denseblock4.denselayer15.norm1.bias \t torch.Size([960])\n",
      "features.denseblock4.denselayer15.norm1.running_mean \t torch.Size([960])\n",
      "features.denseblock4.denselayer15.norm1.running_var \t torch.Size([960])\n",
      "features.denseblock4.denselayer15.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer15.conv1.weight \t torch.Size([128, 960, 1, 1])\n",
      "features.denseblock4.denselayer15.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer15.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer15.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer15.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer15.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer15.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.denseblock4.denselayer16.norm1.weight \t torch.Size([992])\n",
      "features.denseblock4.denselayer16.norm1.bias \t torch.Size([992])\n",
      "features.denseblock4.denselayer16.norm1.running_mean \t torch.Size([992])\n",
      "features.denseblock4.denselayer16.norm1.running_var \t torch.Size([992])\n",
      "features.denseblock4.denselayer16.norm1.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer16.conv1.weight \t torch.Size([128, 992, 1, 1])\n",
      "features.denseblock4.denselayer16.norm2.weight \t torch.Size([128])\n",
      "features.denseblock4.denselayer16.norm2.bias \t torch.Size([128])\n",
      "features.denseblock4.denselayer16.norm2.running_mean \t torch.Size([128])\n",
      "features.denseblock4.denselayer16.norm2.running_var \t torch.Size([128])\n",
      "features.denseblock4.denselayer16.norm2.num_batches_tracked \t torch.Size([])\n",
      "features.denseblock4.denselayer16.conv2.weight \t torch.Size([32, 128, 3, 3])\n",
      "features.norm5.weight \t torch.Size([1024])\n",
      "features.norm5.bias \t torch.Size([1024])\n",
      "features.norm5.running_mean \t torch.Size([1024])\n",
      "features.norm5.running_var \t torch.Size([1024])\n",
      "features.norm5.num_batches_tracked \t torch.Size([])\n",
      "dense.weight \t torch.Size([1024, 50176])\n",
      "dense.bias \t torch.Size([1024])\n",
      "dense2.weight \t torch.Size([1024, 1024])\n",
      "dense2.bias \t torch.Size([1024])\n",
      "dense3.weight \t torch.Size([1024, 1024])\n",
      "dense3.bias \t torch.Size([1024])\n",
      "dense4.weight \t torch.Size([1024, 1024])\n",
      "dense4.bias \t torch.Size([1024])\n",
      "dense5.weight \t torch.Size([1024, 1024])\n",
      "dense5.bias \t torch.Size([1024])\n",
      "classifier.weight \t torch.Size([1, 1024])\n",
      "classifier.bias \t torch.Size([1])\n"
     ]
    }
   ],
   "source": [
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
    "model = Classifier()\n",
    "\n",
    "model.to(device)\n",
    "\n",
    "for param_tensor in model.state_dict():\n",
    "    print(param_tensor, \"\\t\", model.state_dict()[param_tensor].size())\n",
    "    \n",
    "model_save_name = 'model2.pt'\n",
    "\n",
    "criterion = nn.MSELoss()  \n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=0.001)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Evaluate model on validation set from last saved model checkpoint.\n",
    "checkpoint = torch.load(model_save_name)\n",
    "def evaluate(model, dataloader):\n",
    "    model.eval()\n",
    "    losses = []\n",
    "    with torch.no_grad():\n",
    "        torch.manual_seed(2)\n",
    "        for i, data in enumerate(dataloader, checkpoint['batch']+1):     # 0 is start index.\n",
    "            dense_output, labels, img_path = data\n",
    "            output = model(dense_output) \n",
    "            output = output.to(device)\n",
    "            labels = labels.to(device)\n",
    "\n",
    "            loss = criterion(output, labels)\n",
    "            losses.append(loss.item())     # extract only loss, as a PyTorch float.\n",
    "\n",
    "\n",
    "    return sum(losses) / len(losses)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: train model on DenseNet embeddings\n",
    "train_losses = []\n",
    "val_losses = []\n",
    "batch_running_losses = []\n",
    "num_epochs = 8       \n",
    "\n",
    "for epoch in range(num_epochs):\n",
    "    start_epoch_time = time.time()\n",
    "    model.train()  \n",
    "\n",
    "    torch.manual_seed(2)        \n",
    "    for i, data in enumerate(train_loader, 0):      # 0 is start index\n",
    "        dense_output, labels, img_path = data\n",
    "        dense_output.to(device)\n",
    "        labels.to(device)\n",
    "        output = model(dense_output)   \n",
    "        \n",
    "        optimizer.zero_grad()\n",
    "\n",
    "        training_loss = criterion(output, labels)\n",
    "        train_losses.append(training_loss.item())\n",
    "\n",
    "        # Save model progress for each batch.\n",
    "        torch.save({\n",
    "            'epoch': epoch,\n",
    "            'model_state_dict': model.state_dict(),\n",
    "            'optimizer_state_dict': optimizer.state_dict(),\n",
    "            'loss': train_losses\n",
    "            }, model_save_name)\n",
    "\n",
    "        print(\"Model snapshot saved!\")\n",
    "\n",
    "        with open(\"train_ALL3.txt\", \"a\") as f:\n",
    "                f.write(str(training_loss.item()))\n",
    "                f.write('\\n')\n",
    "\n",
    "        training_loss.backward()     # backward pass\n",
    "\n",
    "        optimizer.step()             # step\n",
    "        \n",
    "        val_loss = evaluate(model, val_loader)\n",
    "        val_losses.append(val_loss)\n",
    "        with open(\"val_ALL3.txt\", \"a\") as f2:    \n",
    "            f2.write(str(val_loss))\n",
    "            f2.write('\\n')\n",
    "\n",
    "        model.train()\n",
    "\n",
    "\n",
    "    end_epoch_time = time.time()  # End timer for epoch\n",
    "    epoch_time = end_epoch_time - start_epoch_time\n",
    "    print(f\"Epoch {epoch+1} took {epoch_time} seconds\")\n",
    "    \n",
    "\n",
    "print(\"Done!\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Keep running interrupted training session where it last exited.\n",
    "\n",
    "checkpoint = torch.load(model_save_name)\n",
    "\n",
    "# Load model and optimizer state\n",
    "model.load_state_dict(checkpoint['model_state_dict'])\n",
    "optimizer.load_state_dict(checkpoint['optimizer_state_dict'])\n",
    "\n",
    "for epoch in range(checkpoint['epoch'] + 1, num_epochs):\n",
    "    torch.manual_seed(2) \n",
    "    for i, data in enumerate(train_loader, checkpoint['batch']+1):          \n",
    "        output, labels, img_path = data[0], data[1], data[2]\n",
    "        output.to(device)\n",
    "        labels.to(device)\n",
    "        optimizer.zero_grad()\n",
    "\n",
    "        training_loss = criterion(output, labels)\n",
    "        train_losses.append(training_loss.item())\n",
    "    \n",
    "        torch.save({\n",
    "            'batch':i,\n",
    "            'epoch': epoch,\n",
    "            'model_state_dict': model.state_dict(),\n",
    "            'optimizer_state_dict': optimizer.state_dict(),\n",
    "            'loss': train_losses}, model_save_name)\n",
    "\n",
    "        print(\"Model snapshot saved!\")\n",
    "        \n",
    "        print(\"Model snapshot saved!\")\n",
    "\n",
    "        with open(\"train_ALL3.txt\", \"a\") as f:\n",
    "                f.write(str(training_loss.item()))\n",
    "                f.write('\\n')\n",
    "\n",
    "        training_loss.backward()     # backward pass\n",
    "\n",
    "        optimizer.step()             # step\n",
    "        \n",
    "        val_loss = evaluate(model, val_loader)\n",
    "        val_losses.append(val_loss)\n",
    "        with open(\"val_ALL3.txt\", \"a\") as f2:\n",
    "            f2.write(str(val_loss))\n",
    "            f2.write('\\n')\n",
    "\n",
    "        model.train()\n",
    "        \n",
    "    end_epoch_time = time.time()  # End timer for epoch\n",
    "    epoch_time = end_epoch_time - start_epoch_time\n",
    "    print(f\"Epoch {epoch+1} took {epoch_time} seconds\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Test, not used model.\n",
    "# Loss trend plots\n",
    "\n",
    "losses = []\n",
    "validation = []\n",
    "with open (\"train_ALL3.txt\", 'r') as f:\n",
    "    for line in f: \n",
    "        losses.append(float(line))\n",
    "        \n",
    "\n",
    "with open(\"val_ALL3.txt\", \"r\") as f2:\n",
    "    for line2 in f2:\n",
    "        validation.append(float(line2))\n",
    "        \n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.plot(losses[0:46], label='Train Data')    # We have to make sure to put an index step here that correlates to validation data indices.\n",
    "plt.plot(validation, label= \"Validation data\")\n",
    "plt.xlabel('Batch')\n",
    "plt.ylabel('Loss')\n",
    "plt.ylim(0, 20)\n",
    "plt.title('Loss Trend, LR=0.001')\n",
    "plt.legend()\n",
    "plt.grid(True)\n",
    "plt.show()\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "13\n"
     ]
    }
   ],
   "source": [
    "checkpoint = torch.load(model_save_name)\n",
    "print(checkpoint['batch'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Training, saved code from earlier training sessions.\n",
    "train_losses = []\n",
    "val_losses = []\n",
    "batch_running_losses = []\n",
    "num_epochs = 10       # linear time complexity.\n",
    "\n",
    "for epoch in range(num_epochs):\n",
    "    start_epoch_time = time.time()\n",
    "    model.train()   \n",
    "    \n",
    "    torch.manual_seed(2)        \n",
    "    for i, data in enumerate(train_loader, 0):   \n",
    "        # Around 10% of images for first check of loss convergence:   \n",
    "        \"\"\"if i >= 2500:       # 2500 first data points in the shuffled dataset (around 10%)\n",
    "            break \"\"\"\n",
    "        \n",
    "        \n",
    "        inputs, labels, img_path = data\n",
    "        inputs, labels = inputs.to(device), labels.to(device)\n",
    "        optimizer.zero_grad()  \n",
    "        \n",
    "        outputs = model(inputs)     \n",
    "        \n",
    "        training_loss = criterion(outputs, labels)\n",
    "        train_losses.append(training_loss.item()) \n",
    "        \n",
    "        with open(\"train_ALL3.txt\", \"a\") as f:\n",
    "                f.write(str(training_loss.item()))\n",
    "                f.write('\\n')\n",
    "        \n",
    "        training_loss.backward()     # backward pass\n",
    "        \n",
    "        optimizer.step()             # step\n",
    "        \n",
    "        if i % 50 == 0:        \n",
    "            val_loss = evaluate(model, val_loader)\n",
    "            val_losses.append(val_loss)    \n",
    "            with open(\"val_ALL3.txt\", \"a\") as f2:\n",
    "                f2.write(str(val_loss))\n",
    "                f2.write('\\n')\n",
    "                \n",
    "        model.train()\n",
    "        \n",
    "        torch.save({\n",
    "            'epoch': epoch,\n",
    "            'model_state_dict': model.state_dict(),\n",
    "            'optimizer_state_dict': optimizer.state_dict(),\n",
    "            'loss': train_losses\n",
    "            }, model_save_name)\n",
    "    \n",
    "        print(\"Model snapshot saved!\")\n",
    "            \n",
    "        \n",
    "    end_epoch_time = time.time()  # End timer for epoch\n",
    "    epoch_time = end_epoch_time - start_epoch_time\n",
    "    print(f\"Epoch {epoch+1} took {epoch_time} seconds\")\n",
    "\n",
    "\n",
    "print(\"Done!\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# All training data, loss plot.\n",
    "validation = []\n",
    "with open('val_ALL2.txt', 'r') as val_file:\n",
    "    for line in val_file:\n",
    "        validation.append(line)\n",
    "\n",
    "window_size = 10\n",
    "moving_avg = np.convolve(train_losses, np.ones(window_size)/window_size, mode='valid')\n",
    "indices = np.arange(window_size//2, len(train_losses) - window_size//2 + 1)\n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.plot(train_losses[0:1000], label='Train Data')\n",
    "plt.xlabel('Data sample')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Loss Trend with Moving Average, LR=0.001')\n",
    "\n",
    "plt.xlabel('Data sample')\n",
    "plt.ylabel('Loss')\n",
    "plt.ylim(0, 20)\n",
    "plt.title('Loss Trend, all training data')\n",
    "plt.legend()\n",
    "plt.grid(True)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Loss trend plots, training and validation data.\n",
    "\n",
    "losses = []\n",
    "validation = []\n",
    "with open (\"train_ALL2.txt\", 'r') as f:\n",
    "    for line in f: \n",
    "        losses.append(float(line))\n",
    "        \n",
    "\n",
    "with open(\"val_ALL2.txt\", \"r\") as f2:\n",
    "    for line2 in f2:\n",
    "        validation.append(float(line2))\n",
    "        \n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.plot(losses[0:40], label='Train Data')\n",
    "plt.plot(validation[0:40], label= \"Validation data\")\n",
    "# plt.xticks(range(0, 50, 5))\n",
    "# plt.yticks(range(0, 12500, 1000))\n",
    "plt.xlabel('Batch')\n",
    "plt.ylabel('Loss')\n",
    "plt.ylim(0, 20)\n",
    "plt.title('Loss Trend, LR=0.001')\n",
    "plt.legend()\n",
    "plt.grid(True)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Following code cells will be used for testing the DenseNet model and analyzing its result, or defining new instances of the model for further analysis or experiments for future implementations. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: test model [work in progress for testing new model instances]\n",
    "\n",
    "def test_model(model, dataloader):\n",
    "    model.eval()\n",
    "    predictions = []\n",
    "    true_labels = []\n",
    "    image_paths = []\n",
    "    with torch.no_grad():\n",
    "        for data in dataloader:              \n",
    "            inputs, labels, img_paths = data     \n",
    "            inputs, labels = inputs.to(device), labels.to(device)\n",
    "            outputs = model(inputs)\n",
    "            predictions.extend(outputs.cpu().numpy()) \n",
    "            true_labels.extend(labels.cpu().numpy())\n",
    "            image_paths.extend(img_paths)\n",
    "\n",
    "    predictions = np.array(predictions)\n",
    "    true_labels = np.array(true_labels)\n",
    "    image_paths = np.array(image_paths)\n",
    "\n",
    "    return predictions, true_labels, image_paths"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions, true_labels, image_paths = test_model(model, test_loader)\n",
    "test_df = pd.DataFrame({'Image_Path': image_paths, 'Ground_Truth': true_labels, 'Inference': predictions}) # plocka ut bara komp_nivå, film och crop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "StandardScaler()\n",
      "tensor(1.0821)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'inversed = scaler.inverse_transform(scaled)\\nprint(inversed)'"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Test with scaled data.\n",
    "scaler = test_data.scaler\n",
    "\n",
    "print(scaler)\n",
    "\n",
    "scaled = []\n",
    "for i in range(len(test_data)):\n",
    "    scaled.append(test_data[i][1])\n",
    "    \n",
    "print(scaled[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[97.4274, 97.4274, 97.4274,  ...,  7.7205,  3.8588, 23.8076]])"
      ]
     },
     "execution_count": 97,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "inversed = scaler.inverse_transform([scaled])\n",
    "inversed_tensor = torch.tensor(inversed, dtype=torch.float32)\n",
    "\n",
    "ImagesAndLabelsDataset.inversed = property(lambda self: inversed_tensor)\n",
    "test_data.inversed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                              Image_Path  Ground_Truth  \\\n",
      "0      /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...      1.082086   \n",
      "1      /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...      1.082086   \n",
      "2      /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...      1.082086   \n",
      "3      /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...      1.082086   \n",
      "4      /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...      1.082086   \n",
      "...                                                  ...           ...   \n",
      "17755  /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...     -1.804801   \n",
      "17756  /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...     -1.075653   \n",
      "17757  /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...     -1.867215   \n",
      "17758  /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...     -1.994176   \n",
      "17759  /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...     -1.338318   \n",
      "\n",
      "       Inference Position Compression_Level                  Movie_Name  \n",
      "0      -0.224274   Corner                23        agent-cody-banks.mp4  \n",
      "1      -0.224289   Corner                23        agent-cody-banks.mp4  \n",
      "2      -0.224289   Corner                23        agent-cody-banks.mp4  \n",
      "3      -0.224289   Corner                23        agent-cody-banks.mp4  \n",
      "4      -0.224289     Edge                23        agent-cody-banks.mp4  \n",
      "...          ...      ...               ...                         ...  \n",
      "17755  -1.618250     Edge                51  the-sun-is-also-a-star.mp4  \n",
      "17756  -0.924868     Edge                51  the-sun-is-also-a-star.mp4  \n",
      "17757  -1.868219     Edge                51  the-sun-is-also-a-star.mp4  \n",
      "17758  -1.834139   Center                51  the-sun-is-also-a-star.mp4  \n",
      "17759  -1.394831     Edge                51  the-sun-is-also-a-star.mp4  \n",
      "\n",
      "[17760 rows x 6 columns]\n"
     ]
    }
   ],
   "source": [
    "# TODO: Pick out neccessary information about predictions and save to CSV file.\n",
    "\n",
    "centers = test_df[test_df[\"Image_Path\"].str.contains(\"center\")]\n",
    "corners = test_df[(test_df[\"Image_Path\"].str.contains(\"bottom\") | test_df[\"Image_Path\"].str.contains(\"top\")) & (test_df[\"Image_Path\"].str.contains(\"left\") | test_df[\"Image_Path\"].str.contains(\"right\"))]\n",
    "edges = test_df[(test_df[\"Image_Path\"].str.contains(\"bottom\") | test_df[\"Image_Path\"].str.contains(\"top\") | test_df[\"Image_Path\"].str.contains(\"right\") | test_df[\"Image_Path\"].str.contains(\"left\"))] \n",
    "\n",
    "test_df['Position'] = 'Other'   \n",
    "test_df.loc[test_df.index.isin(edges.index), 'Position'] = 'Edge'\n",
    "test_df.loc[test_df.index.isin(corners.index), 'Position'] = 'Corner'\n",
    "test_df.loc[test_df.index.isin(centers.index), 'Position'] = 'Center'\n",
    "\n",
    "test_df['Compression_Level'] = test_df['Image_Path'].apply(lambda x: x.split('_')[8].split('_frame')[0])\n",
    "test_df['Movie_Name'] = test_df['Image_Path'].apply(lambda x: x.split('_')[9].split('_frame')[0])\n",
    "\n",
    "print(test_df)\n",
    "\n",
    "test_df.to_csv('inference_ALL2.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                              Image_Path  Ground_Truth  \\\n",
      "0      /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...     97.427361   \n",
      "1      /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...     97.427361   \n",
      "2      /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...     97.427361   \n",
      "3      /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...     97.427361   \n",
      "4      /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...     97.427361   \n",
      "...                                                  ...           ...   \n",
      "17755  /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...      9.618851   \n",
      "17756  /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...     31.796865   \n",
      "17757  /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...      7.720450   \n",
      "17758  /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...      3.858760   \n",
      "17759  /Users/kajsalidin/Desktop/nr-vqa-vmaf/images_T...     23.807552   \n",
      "\n",
      "        Inference Position Compression_Level Movie_Name  \n",
      "0      106.869057   Corner        compressed         23  \n",
      "1      106.869057   Corner        compressed         23  \n",
      "2      106.869057   Corner        compressed         23  \n",
      "3      106.869057   Corner        compressed         23  \n",
      "4      106.869057     Edge        compressed         23  \n",
      "...           ...      ...               ...        ...  \n",
      "17755   26.242207     Edge        compressed         51  \n",
      "17756   85.969421     Edge        compressed         51  \n",
      "17757   55.181141     Edge        compressed         51  \n",
      "17758   16.115189   Center        compressed         51  \n",
      "17759   63.360710     Edge        compressed         51  \n",
      "\n",
      "[17760 rows x 6 columns]\n"
     ]
    }
   ],
   "source": [
    "with open(\"inference.pkl\",\"rb\") as f:\n",
    "    obj = pickle.load(f)\n",
    "    print(obj)\n",
    "    \n",
    "# Add to the dictionary object \n",
    "obj[\"new\"]=test_df.to_pickle\n",
    "    \n",
    "with open(\"inference.pkl\",\"wb\") as f:\n",
    "    pickle.dump(obj, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.0820857\n",
      "97.427361\n",
      "1.0820857\n",
      "97.427361\n",
      "0.224937\n",
      "71.35604\n",
      "0.64692056\n",
      "84.191232\n",
      "0.87515265\n",
      "91.133216\n"
     ]
    }
   ],
   "source": [
    "# Double-checking that the ground truths values are correct.\n",
    "\n",
    "csv_file_path = \"inference_ALL2.csv\"\n",
    "csv_file_path2 = \"vmaf_values_TEST_ALL.csv\"\n",
    "vmaf_from_inference= []\n",
    "vmaf_from_gt = []\n",
    "\n",
    "with open(csv_file_path, newline='') as csvfile:\n",
    "    csvfile.readline()\n",
    "    csv_reader = csv.reader(csvfile)\n",
    "    for row in csv_reader:\n",
    "        vmaf_score = float(row[1])\n",
    "        vmaf_from_inference.append(vmaf_score)\n",
    "        \n",
    "vmaf_scores2 = []\n",
    "with open(csv_file_path2, newline='') as csvfile:\n",
    "    csvfile.readline()\n",
    "    csv_reader = csv.reader(csvfile)\n",
    "    for row in csv_reader:\n",
    "        vmaf_score = float(row[0])\n",
    "        vmaf_from_gt.append(vmaf_score)\n",
    "        \n",
    "print(vmaf_from_inference[0])      \n",
    "print(vmaf_from_gt[0])\n",
    "\n",
    "print(vmaf_from_inference[1])      \n",
    "print(vmaf_from_gt[1])\n",
    "\n",
    "print(vmaf_from_inference[100])      \n",
    "print(vmaf_from_gt[100])\n",
    "\n",
    "print(vmaf_from_inference[1200])      \n",
    "print(vmaf_from_gt[1200])\n",
    "\n",
    "print(vmaf_from_inference[5000])      \n",
    "print(vmaf_from_gt[5000])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "r_squared_values = []\n",
    "ground_truths = df[\"Ground_Truth\"]\n",
    "predictions = df[\"Inference\"]\n",
    "\n",
    "r_squared = r2_score(ground_truths, predictions)\n",
    "\n",
    "print(\"R-squared score on the entire test set:\", r_squared)\n",
    "\n",
    "std_dev = np.std(ground_truths - predictions)\n",
    "print(\"Standard deviation of the entire test set\", std_dev)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# When testing model on images_TEST_comp_crop_ALL, the compression levels are 23, 37 and 51 instead.\n",
    "\n",
    "corners = test_df.loc[test_df['Position'] == 'Corner']\n",
    "corners_compression_23 = corners.loc[corners['Compression_Level'] == '23']\n",
    "corners_compression_50 = corners.loc[corners['Compression_Level'] == '50']\n",
    "corners_compression_51 = corners.loc[corners['Compression_Level'] == '51']\n",
    "\n",
    "centers = test_df.loc[test_df['Position'] == 'Center']\n",
    "centers_compression_23 = centers.loc[centers['Compression_Level'] == '23']\n",
    "centers_compression_50 = centers.loc[centers['Compression_Level'] == '50']\n",
    "centers_compression_51 = centers.loc[centers['Compression_Level'] == '51']\n",
    "\n",
    "edges = test_df.loc[test_df['Position'] == 'Edge']\n",
    "edges_compression_23 = edges.loc[edges['Compression_Level'] == '23']\n",
    "edges_compression_50 = edges.loc[edges['Compression_Level'] == '50']\n",
    "edges_compression_51 = edges.loc[edges['Compression_Level'] == '51']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(len(corners_compression_23))\n",
    "print(len(corners_compression_50))\n",
    "print(len(corners_compression_51))\n",
    "\n",
    "print()\n",
    "print(len(centers_compression_23))\n",
    "print(len(centers_compression_50))\n",
    "print(len(centers_compression_51))\n",
    "\n",
    "print()\n",
    "print(len(edges_compression_23))\n",
    "print(len(edges_compression_50))\n",
    "print(len(edges_compression_51))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Calculate absolute differences between true and predicted values.\n",
    "\n",
    "def abs_diff(gt, pred):      \n",
    "    diff = gt - pred\n",
    "    return abs(diff)        \n",
    "\n",
    "abs_diffs_corners_23 = []\n",
    "for true, pred in zip(corners_compression_23['Ground_Truth'], corners_compression_23['Inference']): \n",
    "    abs_diffs_corners_23.append(abs_diff(true, pred))\n",
    "\n",
    "std_corners_23 = np.std(corners_compression_23['Inference'])\n",
    "    \n",
    "abs_diffs_corners_50 = []\n",
    "for true, pred in zip(corners_compression_50['Ground_Truth'], corners_compression_50['Inference']): \n",
    "    abs_diffs_corners_50.append(abs_diff(true, pred))\n",
    "\n",
    "std_corners_50 = np.std(corners_compression_50['Inference'])\n",
    "    \n",
    "abs_diffs_corners_51 = []\n",
    "for true, pred in zip(corners_compression_51['Ground_Truth'], corners_compression_51['Inference']): \n",
    "    abs_diffs_corners_51.append(abs_diff(true, pred))\n",
    "\n",
    "std_corners_51 = np.std(corners_compression_51['Inference'])\n",
    "\n",
    "abs_diffs_centers_23 = []\n",
    "for true, pred in zip(centers_compression_23['Ground_Truth'], centers_compression_23['Inference']): \n",
    "    abs_diffs_centers_23.append(abs_diff(true, pred))\n",
    "\n",
    "std_centers_23 = np.std(centers_compression_23['Inference'])\n",
    "    \n",
    "abs_diffs_centers_50 = []\n",
    "for true, pred in zip(centers_compression_50['Ground_Truth'], centers_compression_50['Inference']): \n",
    "    abs_diffs_centers_50.append(abs_diff(true, pred))\n",
    "\n",
    "std_centers_50 = np.std(centers_compression_50['Inference'])\n",
    "    \n",
    "abs_diffs_centers_51 = []\n",
    "for true, pred in zip(centers_compression_51['Ground_Truth'], centers_compression_51['Inference']): \n",
    "    abs_diffs_centers_51.append(abs_diff(true, pred))\n",
    "\n",
    "std_centers_51 = np.std(centers_compression_51['Inference'])\n",
    "    \n",
    "abs_diffs_edges_23 = []\n",
    "for true, pred in zip(edges_compression_23['Ground_Truth'], edges_compression_23['Inference']): \n",
    "    abs_diffs_edges_23.append(abs_diff(true, pred))\n",
    "\n",
    "std_edges_23 = np.std(edges_compression_23['Inference'])\n",
    "    \n",
    "abs_diffs_edges_50 = []\n",
    "for true, pred in zip(edges_compression_50['Ground_Truth'], edges_compression_50['Inference']): \n",
    "    abs_diffs_edges_50.append(abs_diff(true, pred))\n",
    "\n",
    "std_edges_50 = np.std(edges_compression_50['Inference'])\n",
    "    \n",
    "abs_diffs_edges_51 = []\n",
    "for true, pred in zip(edges_compression_51['Ground_Truth'], edges_compression_51['Inference']): \n",
    "    abs_diffs_edges_51.append(abs_diff(true, pred))\n",
    "\n",
    "std_edges_51 = np.std(edges_compression_51['Inference'])\n",
    "   \n",
    "    \n",
    "print(sum(abs_diffs_corners_23)/ len(abs_diffs_corners_23))\n",
    "print(std_corners_23)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "categories = [\"Low comp corner\", \"Medium comp corner\", \"Highest comp corner\", \"Low comp center\", \"Medium comp center\", \"Highest comp center\", \"Low comp edge\", \"Medium comp edge\", \"Highest comp edge\"]\n",
    "avg_diffs = [sum(abs_diffs_corners_23)/len(abs_diffs_corners_23), sum(abs_diffs_corners_50)/len(abs_diffs_corners_50), sum(abs_diffs_corners_51)/len(abs_diffs_corners_51), sum(abs_diffs_centers_23)/len(abs_diffs_centers_23), sum(abs_diffs_centers_50)/len(abs_diffs_centers_50), sum(abs_diffs_centers_51) / len(abs_diffs_centers_51), sum(abs_diffs_edges_23)/ len(abs_diffs_edges_23), sum(abs_diffs_edges_50)/len(abs_diffs_edges_50), sum(abs_diffs_edges_51)/len(abs_diffs_edges_51)]\n",
    "colors = [\"#ea9999\", \"#e06666\", \"#ea4c4c\", \"#b6d7a8\", \"#93c47d\", \"#6aa84f\", \"#9fc5e8\", \"#6fa8dc\", \"#3d85c6\"]\n",
    "\n",
    "plt.bar(categories, avg_diffs, color = colors)\n",
    "plt.xticks(rotation='vertical')\n",
    "plt.yticks(range(0, 25, 3))\n",
    "plt.title(\"Mean absolute error based on image category\")\n",
    "plt.grid(alpha=0.2)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "categories = [\"Low comp corner\", \"High comp corner\", \"Highest comp corner\", \"Low comp center\", \"High comp center\", \"Highest comp center\", \"Low comp edge\", \"High comp edge\", \"Highest comp edge\"]\n",
    "stds = [std_corners_23, std_corners_50, std_corners_51, std_centers_23, std_centers_50, std_centers_51, std_edges_23, std_edges_50, std_edges_51]\n",
    "colors = [\"#ea9999\", \"#e06666\", \"#ea4c4c\", \"#b6d7a8\", \"#93c47d\", \"#6aa84f\", \"#9fc5e8\", \"#6fa8dc\", \"#3d85c6\"]\n",
    "\n",
    "plt.bar(categories, stds, color = colors)\n",
    "plt.xticks(rotation='vertical')\n",
    "plt.title(\"Standard deviation of predicted values based on image category\")\n",
    "plt.grid(alpha=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "categories = [\"Low comp corner\", \"High comp corner\", \"Highest comp corner\", \"Low comp center\", \"High comp center\", \"Highest comp center\", \"Low comp edge\", \"High comp edge\", \"Highest comp edge\"]\n",
    "stds_true = [np.std(corners_compression_23['Ground_Truth']), np.std(corners_compression_50['Ground_Truth']), np.std(corners_compression_51['Ground_Truth']), np.std(centers_compression_23['Ground_Truth']), np.std(centers_compression_50['Ground_Truth']), np.std(centers_compression_51['Ground_Truth']), np.std(edges_compression_23['Ground_Truth']), np.std(edges_compression_50['Ground_Truth']), np.std(edges_compression_51['Ground_Truth'])]\n",
    "colors = [\"#ea9999\", \"#e06666\", \"#ea4c4c\", \"#b6d7a8\", \"#93c47d\", \"#6aa84f\", \"#9fc5e8\", \"#6fa8dc\", \"#3d85c6\"]\n",
    "\n",
    "plt.bar(categories, stds_true, color = colors)\n",
    "plt.xticks(rotation='vertical')\n",
    "plt.title(\"Standard deviation of ground truth values based on image category\")\n",
    "plt.grid(alpha=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "categories = [\"Low comp corner\", \"High comp corner\", \"Highest comp corner\", \"Low comp center\", \"High comp center\", \"Highest comp center\", \"Low comp edge\", \"High comp edge\", \"Highest comp edge\"]\n",
    "avg_vmaf_true = [sum(corners_compression_23['Ground_Truth'])/len(corners_compression_23['Ground_Truth']), sum(corners_compression_50['Ground_Truth'])/ len(corners_compression_50['Ground_Truth']), sum(corners_compression_51['Ground_Truth'])/len(corners_compression_51['Ground_Truth']), sum(centers_compression_23['Ground_Truth'])/len(centers_compression_23['Ground_Truth']), sum(centers_compression_50['Ground_Truth'])/len(centers_compression_50['Ground_Truth']), sum(centers_compression_51['Ground_Truth'])/len(centers_compression_51['Ground_Truth']), sum(edges_compression_23['Ground_Truth'])/len(edges_compression_23['Ground_Truth']), sum(edges_compression_50['Ground_Truth'])/len(edges_compression_50['Ground_Truth']), sum(edges_compression_51['Ground_Truth'])/len(edges_compression_51['Ground_Truth'])]\n",
    "avg_vmaf_pred = [sum(corners_compression_23['Inference'])/len(corners_compression_23['Inference']), sum(corners_compression_50['Inference'])/ len(corners_compression_50['Inference']), sum(corners_compression_51['Inference'])/len(corners_compression_51['Inference']), sum(centers_compression_23['Inference'])/len(centers_compression_23['Inference']), sum(centers_compression_50['Inference'])/len(centers_compression_50['Inference']), sum(centers_compression_51['Inference'])/len(centers_compression_51['Inference']), sum(edges_compression_23['Inference'])/len(edges_compression_23['Inference']), sum(edges_compression_50['Inference'])/len(edges_compression_50['Inference']), sum(edges_compression_51['Inference'])/len(edges_compression_51['Inference'])]\n",
    "colors = [\"#ea9999\", \"#e06666\", \"#ea4c4c\", \"#b6d7a8\", \"#93c47d\", \"#6aa84f\", \"#9fc5e8\", \"#6fa8dc\", \"#3d85c6\"]\n",
    "\n",
    "plt.bar(categories, avg_vmaf_true, color = colors)\n",
    "plt.title(\"Average VMAF, true labels\")\n",
    "plt.ylabel('VMAF')\n",
    "plt.xticks(rotation='vertical')\n",
    "plt.grid(alpha=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.bar(categories, avg_vmaf_pred, color = colors)\n",
    "plt.title(\"Average VMAF, predicted labels\")\n",
    "plt.xticks(rotation='vertical')\n",
    "plt.grid(alpha=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "abs_diffs_movies = []\n",
    "\n",
    "average_ground_truth = test_df.groupby(['Movie_Name'])['Ground_Truth'].mean().reset_index()\n",
    "average_vmaf_pred = test_df.groupby(['Movie_Name'])['Inference'].mean().reset_index()\n",
    "\n",
    "abs_diffs_movies = []\n",
    "for movie in test_df['Movie_Name'].unique():\n",
    "    abs_diff = abs(test_df.loc[test_df['Movie_Name'] == movie, 'Ground_Truth'] - test_df.loc[test_df['Movie_Name'] == movie, 'Inference'])\n",
    "    abs_diffs_movies.append(abs_diff.mean())\n",
    "    \n",
    "print(\"MAE loss based on movie name is\", sum(abs_diffs_movies)/len(abs_diffs_movies))\n",
    "\n",
    "r_squared = r2_score(test_df['Ground_Truth'], test_df['Inference'])\n",
    "print(\"R^2 score of model performance, sorted by movie is\", r_squared)\n",
    "\n",
    "plt.plot(test_df['Inference'][20:100], label=\"Average ground truth VMAF\")\n",
    "plt.plot(test_df['Ground_Truth'][20:100], label=\"Average predicted VMAF\")\n",
    "plt.xticks(rotation=45, ha='right', rotation_mode='anchor')  \n",
    "plt.legend(fontsize='small')\n",
    "plt.title('Average VMAF values based on movie: true and predicted values')\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Group by Movie_Name, Compression_Level, and Position and calculate average Inference values.\n",
    "grouped_df = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Inference'].mean().reset_index()\n",
    "\n",
    "# Define categories for X-axis labels.\n",
    "categories = [f\"{movie} - {compression} - {position}\" for movie, compression, position in zip(grouped_df['Movie_Name'], grouped_df['Compression_Level'].replace('75', '51'), grouped_df['Position'])]\n",
    "colors = []\n",
    "for cat in categories:\n",
    "    if \"Corner\" in cat:\n",
    "        if \"23\" in cat:\n",
    "            colors.append(\"#f5adad\")\n",
    "            \n",
    "        elif \"50\" in cat:\n",
    "            colors.append(\"#ff0000\")\n",
    "            \n",
    "        elif \"51\" in cat:\n",
    "            colors.append(\"#a41919\")\n",
    "            \n",
    "    elif \"Center\" in cat:\n",
    "        if \"23\" in cat:\n",
    "            colors.append(\"#b2ebb2\")\n",
    "            \n",
    "        elif \"50\" in cat:\n",
    "            colors.append(\"#099f09\")\n",
    "            \n",
    "        elif \"51\" in cat:\n",
    "            colors.append(\"#056c05\")\n",
    "          \n",
    "    elif \"Edge\" in cat:\n",
    "        if \"23\" in cat:\n",
    "            colors.append(\"#c2c2fe\")\n",
    "            \n",
    "        elif \"50\" in cat:\n",
    "            colors.append(\"#7171e2\")\n",
    "            \n",
    "        elif \"51\" in cat:\n",
    "            colors.append(\"#0b0b95\")  \n",
    "    \n",
    "            \n",
    "# print(colors)\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.bar(categories, grouped_df['Inference'], color=colors)\n",
    "plt.title('Average VMAF Score based on movie, compression level, and position')\n",
    "plt.ylabel('Avg VMAF (pred)')\n",
    "plt.xticks(rotation=90, fontsize=5)\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Function for simplified color settings.\n",
    "grouped_df = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Ground_Truth'].mean().reset_index()\n",
    "\n",
    "# Define categories for X-axis labels.\n",
    "categories = [f\"{movie} - {compression} - {position}\" for movie, compression, position in zip(grouped_df['Movie_Name'], grouped_df['Compression_Level'], grouped_df['Position'])]\n",
    "\n",
    "def colors(categories):\n",
    "    colors = []\n",
    "    for cat in categories:\n",
    "        if \"Corner\" in cat:\n",
    "            if \"23\" in cat:\n",
    "                colors.append(\"#f5adad\")\n",
    "                \n",
    "            elif \"50\" in cat:\n",
    "                colors.append(\"#ff0000\")\n",
    "                \n",
    "            elif \"51\" in cat:\n",
    "                colors.append(\"#a41919\")\n",
    "                \n",
    "        elif \"Center\" in cat:\n",
    "            if \"23\" in cat:\n",
    "                colors.append(\"#b2ebb2\")\n",
    "                \n",
    "            elif \"50\" in cat:\n",
    "                colors.append(\"#099f09\")\n",
    "                \n",
    "            elif \"51\" in cat:\n",
    "                colors.append(\"#056c05\")\n",
    "            \n",
    "        elif \"Edge\" in cat:\n",
    "            if \"23\" in cat:\n",
    "                colors.append(\"#c2c2fe\")\n",
    "                \n",
    "            elif \"50\" in cat:\n",
    "                colors.append(\"#7171e2\")\n",
    "                \n",
    "            elif \"51\" in cat:\n",
    "                colors.append(\"#0b0b95\")  \n",
    "        \n",
    "    return colors\n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.bar(categories, grouped_df['Ground_Truth'], color=colors(categories))\n",
    "plt.title('Average VMAF Score based on movie, compression level, and position')\n",
    "plt.ylabel('Avg VMAF (true)')\n",
    "plt.xticks(rotation=90, fontsize=5)\n",
    "plt.ylim(0, 100)\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(min(round(grouped_df['Ground_Truth'])))\n",
    "print(min(np.ceil(grouped_df['Ground_Truth'])))\n",
    "print(min(grouped_df['Ground_Truth']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.hist(round(test_df['Ground_Truth']), bins=int(max(round(test_df['Ground_Truth']))) - int(min(round(test_df['Ground_Truth']))) + 1, color='cornflowerblue')\n",
    "plt.title('Distribution of ground truth VMAF values')\n",
    "plt.ylabel('Quantity')\n",
    "plt.xticks(range(0, 100, 10))\n",
    "plt.xlabel('VMAF')\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.hist(round(test_df['Inference']), bins=int(max(round(test_df['Inference']))) - int(min(round(test_df['Inference']))) + 1, color='cadetblue')\n",
    "plt.title('Distribution of predicted VMAF values')\n",
    "plt.ylabel('Quantity')\n",
    "plt.xticks(range(0, int(max(round(test_df['Inference']))), 10))\n",
    "plt.xlabel('VMAF (pred)')\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.hist(round(test_df['Inference']), bins=int(max(round(test_df['Inference']))) - int(min(round(test_df['Inference']))) + 1, color='cadetblue', density=True)\n",
    "plt.title('Probability distribution of predicted VMAF values')\n",
    "plt.ylabel('Probability')\n",
    "plt.xticks(range(0, 105, 10))\n",
    "plt.xlim([0, 100])\n",
    "plt.xlabel('VMAF (pred)')\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "grouped_df = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Inference'].std().reset_index()\n",
    "categories = [f\"{movie} - {compression} - {position}\" for movie, compression, position in zip(grouped_df['Movie_Name'], grouped_df['Compression_Level'], grouped_df['Position'])]\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.bar(categories, grouped_df['Inference'], color=colors(categories))\n",
    "plt.title('Standard deviation per movie, compression level and frame position, predicted distribution')\n",
    "plt.ylabel('Standard deviation (pred)')\n",
    "plt.xticks(rotation=90, fontsize=5)\n",
    "plt.ylim(0, 45)\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "grouped_df1 = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Inference'].mean().reset_index()\n",
    "grouped_df2 = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Ground_Truth'].mean().reset_index()\n",
    "\n",
    "mae = mean_absolute_error([grouped_df2['Ground_Truth']],[grouped_df1['Inference']], multioutput='raw_values')\n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.bar(categories, mae, color=colors(categories))\n",
    "plt.title('Mean absolute error based on movie, compression level, and position')\n",
    "plt.ylabel('MAE')\n",
    "plt.xticks(rotation=90, fontsize=5)\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "rounded_mae = []\n",
    "for e in mae:\n",
    "    rounded_mae.append(round(e))\n",
    "    \n",
    "# print(rounded_mae)\n",
    "\n",
    "plt.hist(rounded_mae, bins=max(rounded_mae)-min(rounded_mae)+1, align='left', rwidth=0.8, color='skyblue')\n",
    "plt.xticks(range(0, max(rounded_mae), 2))\n",
    "\n",
    "plt.xlabel('MAE')\n",
    "plt.ylabel('Quantity')\n",
    "plt.grid(alpha=0.2)\n",
    "plt.title('Distribution of Mean Absolute Errors')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.hist(rounded_mae, bins=max(rounded_mae)-min(rounded_mae)+1, align='left', rwidth=0.8, color='skyblue', density=True)\n",
    "plt.xticks(range(0, max(rounded_mae), 2))\n",
    "\n",
    "plt.xlabel('MAE')\n",
    "plt.ylabel('Probability')\n",
    "plt.grid(alpha=0.2)\n",
    "plt.title('Probability distribution of Mean Absolute Errors')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "mae2 = mean_absolute_error([round(grouped_df1['Inference'])], [round(grouped_df2['Ground_Truth'])], multioutput='raw_values')\n",
    "print(mae2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(len(grouped_df1))\n",
    "print(len(test_df))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "indices = []\n",
    "\n",
    "grouped_df1 = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Inference'].mean().reset_index()\n",
    "grouped_df2 = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Ground_Truth'].mean().reset_index()\n",
    "\n",
    "mae = mean_absolute_error([grouped_df2['Ground_Truth']],[grouped_df1['Inference']], multioutput='raw_values')\n",
    "\n",
    "for i in range(len(grouped_df1['Inference'])):\n",
    "    if mae[i] > 6:\n",
    "        indices.append(i)\n",
    "        \n",
    "categories = [f\"{movie} - {compression} - {position}\" for movie, compression, position in zip(grouped_df1['Movie_Name'][[i for i in indices]], grouped_df1['Compression_Level'][[i for i in indices]], grouped_df1['Position'][[i for i in indices]])]\n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.bar(categories, mae[[i for i in indices]], color=colors(categories))\n",
    "plt.title('Mean absolute error outside of allowed error range (>6)')\n",
    "plt.ylabel('MAE')\n",
    "plt.xticks(rotation=90, fontsize=5)\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "indices = []\n",
    "\n",
    "for i in range(len(grouped_df1['Inference'])):\n",
    "    if mae[i] <= 6:\n",
    "        indices.append(i)\n",
    "        \n",
    "categories = [f\"{movie} - {compression} - {position}\" for movie, compression, position in zip(grouped_df1['Movie_Name'][[i for i in indices]], grouped_df1['Compression_Level'].replace('75', '51')[[i for i in indices]], grouped_df1['Position'][[i for i in indices]])]\n",
    "\n",
    "\n",
    "print(indices)\n",
    "\n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.bar(categories, mae[[i for i in indices]], color=colors(categories))\n",
    "plt.title('Mean absolute error, allowed error range (<=6)')\n",
    "plt.ylabel('MAE')\n",
    "plt.xticks(rotation=90)\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_df = pd.read_csv('inference.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "mae = mean_absolute_error([test_df['Ground_Truth']], [test_df['Inference']], multioutput='raw_values')\n",
    "print(\"Average error, sorted by movie, compression level and frame location is\", mae.mean())\n",
    "r_squared = r2_score(test_df['Ground_Truth'], test_df['Inference'])\n",
    "print(\"R^2 score of model performance, sorted by movie, compression level and frame location is\", r_squared)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "mae = mean_absolute_error([grouped_df2['Ground_Truth']], [grouped_df1['Inference']], multioutput='raw_values')\n",
    "print(\"Average error, sorted by movie, compression level and frame location is\", mae.mean())\n",
    "r_squared = r2_score(grouped_df2['Ground_Truth'], grouped_df1['Inference'])\n",
    "print(\"R^2 score of model performance, sorted by movie, compression level and frame location is\", r_squared)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "grouped_df1 = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Inference'].mean().reset_index()\n",
    "grouped_df2 = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Ground_Truth'].mean().reset_index()\n",
    "\n",
    "print(len(grouped_df2['Ground_Truth'].to_numpy()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(len(grouped_df2['Ground_Truth'].to_numpy()))\n",
    "print(len(grouped_df))\n",
    "\n",
    "print(sorted(grouped_df2['Ground_Truth'].to_numpy()))\n",
    "\n",
    "colors = sorted(grouped_df2['Ground_Truth'].to_numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fig, ax = plt.subplots()\n",
    "ax.scatter(grouped_df2['Ground_Truth'].to_numpy(), grouped_df1['Inference'].to_numpy(), c='yellow', edgecolors= \"black\")\n",
    "\n",
    "m, b = np.polyfit(sorted(grouped_df2['Ground_Truth'].to_numpy()), sorted(grouped_df2['Ground_Truth'].to_numpy()), 1)\n",
    "\n",
    "# adding the regression line to the scatter plot.\n",
    "plt.plot(grouped_df2['Ground_Truth'].to_numpy(), m*grouped_df2['Ground_Truth'].to_numpy() + b, color='r')\n",
    "ax.set_xlabel('Actual')\n",
    "ax.set_ylabel('Predicted')\n",
    "# regression line\n",
    "ax.grid(alpha=0.2)\n",
    "plt.yticks(range(20, 120, 10))\n",
    "y_test, y_predicted = grouped_df2['Ground_Truth'].to_numpy().reshape(-1,1), grouped_df1['Inference'].to_numpy().reshape(-1,1)\n",
    "ax.set_title(\"R-squared ≈ {:.3f}\".format(r2_score(grouped_df2['Ground_Truth'].to_numpy(), grouped_df1['Inference'].to_numpy())))\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fig, ax = plt.subplots()\n",
    "ax.scatter(grouped_df2['Ground_Truth'].to_numpy(), grouped_df1['Inference'].to_numpy(), c=\"y\", edgecolors= \"black\")\n",
    "\n",
    "m, b = np.polyfit(grouped_df2['Ground_Truth'], grouped_df1['Inference'].to_numpy(), 1)\n",
    "\n",
    "plt.plot(grouped_df2['Ground_Truth'].to_numpy(), m*grouped_df2['Ground_Truth'].to_numpy() + b, color='r')\n",
    "ax.set_xlabel('Actual')\n",
    "ax.set_ylabel('Predicted')\n",
    "ax.grid(alpha=0.2)\n",
    "y_test, y_predicted = grouped_df2['Ground_Truth'].to_numpy().reshape(-1,1), grouped_df1['Inference'].to_numpy().reshape(-1,1)\n",
    "ax.set_title(\"R-squared ≈ {:.3f}\".format(r2_score(grouped_df2['Ground_Truth'].to_numpy(), grouped_df1['Inference'].to_numpy())))\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "residuals = grouped_df2['Ground_Truth'].to_numpy() - grouped_df1['Inference'].to_numpy()\n",
    "plt.scatter(grouped_df1['Inference'].to_numpy(), residuals)\n",
    "plt.axhline(y=0, color='r', linestyle='--')\n",
    "plt.grid(alpha=0.2)\n",
    "plt.xlabel('Predicted VMAF')\n",
    "plt.ylabel('Residuals')\n",
    "plt.title('Residual plot')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.hist(residuals, bins=10, density=True)\n",
    "plt.grid(alpha=0.2)\n",
    "plt.xlabel(f\"Residuals, Observed mean:{residuals.mean()}\")\n",
    "plt.ylabel(\"Probability\")\n",
    "plt.title(\"Histogram of Residuals\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('vmaf_values_train.csv') as trainfile:\n",
    "    train_dataframe = pd.DataFrame(csv.reader(trainfile, delimiter=\",\"))\n",
    "    train_dataframe.columns= train_dataframe.iloc[0]\n",
    "    train_vmafs = []\n",
    "    train_files = []\n",
    "    for i in range(1, len(train_dataframe)):\n",
    "        train_vmafs.append(float(train_dataframe.iloc[i][0]))\n",
    "        train_files.append(train_dataframe.iloc[i][1])\n",
    "    \n",
    "    \n",
    "with open('vmaf_values_TEST.csv') as testfile:\n",
    "    test_dataframe = pd.DataFrame(csv.reader(testfile, delimiter=\",\"))\n",
    "    test_dataframe.columns= test_dataframe.iloc[0]\n",
    "    test_vmafs = []\n",
    "    test_files = []\n",
    "    for j in range(1, len(test_dataframe)):\n",
    "        test_vmafs.append(float(test_dataframe.iloc[j][0]))\n",
    "        test_files.append(train_dataframe.iloc[i][1])\n",
    "\n",
    "print(np.mean(train_vmafs))       \n",
    "print(np.mean(test_vmafs))  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(len(test_dataframe))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "low_test = test_dataframe[test_dataframe[\"Filename\"].str.contains(\"23\")]\n",
    "high_test = test_dataframe[test_dataframe[\"Filename\"].str.contains(\"50\")]\n",
    "higher_test = test_dataframe[test_dataframe[\"Filename\"].str.contains(\"51\")]\n",
    "\n",
    "low_train = train_dataframe[train_dataframe[\"Filename\"].str.contains(\"23\")]\n",
    "high_train = train_dataframe[train_dataframe[\"Filename\"].str.contains(\"50\")]\n",
    "higher_train = train_dataframe[train_dataframe[\"Filename\"].str.contains(\"51\")]\n",
    "\n",
    "print(len(low_test) / (len(low_test) + len(high_test) + len(higher_test)))\n",
    "print(len(high_test) / (len(low_test) + len(high_test) + len(higher_test)))\n",
    "print(len(higher_test) / (len(low_test) + len(high_test) + len(higher_test)))\n",
    "\n",
    "print()\n",
    "\n",
    "print(len(low_train) / (len(low_train) + len(high_train) + len(higher_train)))\n",
    "print(len(high_train) / (len(low_train) + len(high_train) + len(higher_train)))\n",
    "print(len(higher_train) / (len(low_train) + len(high_train) + len(higher_train)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Acceptable MAE errors are between 0-6.\n",
    "tot_errors = len(mae)\n",
    "acceptable_errors = []\n",
    "non_acceptable_errors = []\n",
    "for e in mae:\n",
    "    if e <= 6:\n",
    "        acceptable_errors.append(e)\n",
    "        \n",
    "    else:\n",
    "        non_acceptable_errors.append(e)\n",
    "\n",
    "print(tot_errors)\n",
    "      \n",
    "print(len(acceptable_errors))\n",
    "\n",
    "print(len(acceptable_errors) / tot_errors)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(grouped_df2['Ground_Truth'].mean())\n",
    "print(grouped_df2['Ground_Truth'].std())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Confidence interval of estimate, try with both confidence level of 95% and 99% (95% is standard)\n",
    "\n",
    "conf = 0.95\n",
    "st.norm.interval(conf, loc=np.mean(grouped_df1['Inference']), scale=st.sem(grouped_df1['Inference'])) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "grouped_df1 = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Inference'].mean().reset_index()\n",
    "grouped_df2 = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Ground_Truth'].mean().reset_index()\n",
    "\n",
    "categories = [f\"{movie} - {compression} - {position}\" for movie, compression, position in zip(grouped_df['Movie_Name'], grouped_df['Compression_Level'].replace('75', '51'), grouped_df['Position'])]\n",
    "mse = mean_squared_error([grouped_df1['Inference']], [grouped_df2['Ground_Truth']], multioutput='raw_values')\n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.bar(categories, mse, color=colors(categories))\n",
    "plt.title('Mean squared error based on movie, compression level, and position')\n",
    "plt.ylabel('MSE')\n",
    "plt.xticks(rotation=90)\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(\"inference_ALL.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "grouped_df_23 = df.where(df['Compression_Level'] == 23).dropna()\n",
    "\n",
    "mae_23 = mean_absolute_error([grouped_df_23['Ground_Truth']], [grouped_df_23['Inference']], multioutput='raw_values')\n",
    "\n",
    "r_squared_23 = r2_score(grouped_df_23['Ground_Truth'], grouped_df_23['Inference'])\n",
    "print(\"R^2 score of model performance, sorted by movie, compression level and frame location is\", r_squared_23)\n",
    "print(\"Average Error is \", mae_23.mean())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train = pd.read_csv('vmaf_values_train_ALL.csv')\n",
    "print(np.mean(df_train['VMAF Score']))\n",
    "print(np.std(df_train['VMAF Score']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "gt_mean = np.mean(df['Ground_Truth'])\n",
    "gt_std = np.std(df['Ground_Truth'])\n",
    "pred_mean = np.mean(df['Inference'])\n",
    "pred_std = np.std(df['Inference'])\n",
    "\n",
    "print(gt_mean)\n",
    "print(gt_std)\n",
    "\n",
    "print(pred_mean)\n",
    "print(pred_std)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Grouped dataframes and statistical measures.\n",
    "grouped_df1 = df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Inference'].mean().reset_index()\n",
    "grouped_df2 = df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Ground_Truth'].mean().reset_index()\n",
    "mae = mean_absolute_error([grouped_df2['Ground_Truth']], [grouped_df1['Inference']], multioutput='raw_values')\n",
    "\n",
    "r_squared = r2_score(grouped_df2['Ground_Truth'], grouped_df1['Inference'])\n",
    "print(\"R^2 score of model performance, sorted by movie, compression level and frame location is\", r_squared)\n",
    "print(\"Average Error is \", mae.mean())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "indices = []\n",
    "\n",
    "for i in range(len(grouped_df1['Inference'])):\n",
    "    if mae[i] <= 6:\n",
    "        indices.append(i)\n",
    "        \n",
    "        \n",
    "categories = [f\"{movie} - {compression} - {position}\" for movie, compression, position in zip(grouped_df1['Movie_Name'][[i for i in indices]], grouped_df1['Compression_Level'].replace('75', '51')[[i for i in indices]], grouped_df1['Position'][[i for i in indices]])]\n",
    "\n",
    "\n",
    "print(indices)\n",
    "\n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.bar(categories, mae[[i for i in indices]], color=colors(categories))\n",
    "plt.title('Mean absolute error, allowed error range (<=6)')\n",
    "plt.ylabel('MAE')\n",
    "plt.xticks(rotation=90, fontsize=5)\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "indices = []\n",
    "\n",
    "for i in range(len(df['Inference'])):\n",
    "    if mae[i] <= 6:\n",
    "        indices.append(i)\n",
    "        \n",
    "        \n",
    "categories = [f\"{movie} - {compression} - {position}\" for movie, compression, position in zip(df['Movie_Name'][[i for i in indices]], df['Compression_Level'][[i for i in indices]], df['Position'][[i for i in indices]])]\n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.bar(categories, mae[[i for i in indices]], color=colors(categories))\n",
    "plt.title('Mean absolute error, allowed error range (<=6)')\n",
    "plt.ylabel('MAE')\n",
    "plt.xticks(rotation=90, fontsize=5)\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "grouped_df1 = df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Inference'].mean().reset_index()\n",
    "grouped_df2 = df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Ground_Truth'].mean().reset_index()\n",
    "\n",
    "print(grouped_df2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fig, ax = plt.subplots()\n",
    "\n",
    "ax.scatter(grouped_df2['Ground_Truth'].to_numpy(), grouped_df1['Inference'].to_numpy(), c='yellow', edgecolors= \"black\")\n",
    "\n",
    "m, b = np.polyfit(sorted(grouped_df2['Ground_Truth'].to_numpy()), sorted(grouped_df2['Ground_Truth'].to_numpy()), 1)\n",
    "\n",
    "plt.plot(grouped_df2['Ground_Truth'].to_numpy(), m*grouped_df2['Ground_Truth'].to_numpy() + b, color='r')\n",
    "\n",
    "plt.grid(alpha=0.2)\n",
    "ax.set_xlabel('Actual')\n",
    "ax.set_ylabel('Predicted')\n",
    "ax.set_title(\"R-squared ≈ {:.3f}\".format(r2_score(grouped_df2['Ground_Truth'].to_numpy(), grouped_df1['Inference'].to_numpy())))\n",
    "\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "residuals = grouped_df2['Ground_Truth'].to_numpy() - grouped_df1['Inference'].to_numpy()\n",
    "plt.scatter(grouped_df1['Inference'].to_numpy(), residuals)\n",
    "plt.axhline(y=0, color='r', linestyle='--')\n",
    "plt.grid(alpha=0.2)\n",
    "plt.xlabel('Predicted VMAF')\n",
    "plt.ylabel('Residuals')\n",
    "plt.title('Residual plot')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "residuals = grouped_df2['Ground_Truth'].to_numpy() - grouped_df1['Inference'].to_numpy()\n",
    "plt.hist(residuals, bins=10, density=True)\n",
    "plt.grid(alpha=0.2)\n",
    "plt.xlabel(f\"Residuals, Observed mean:{residuals.mean()}\")\n",
    "plt.ylabel(\"Probability\")\n",
    "plt.title(\"Histogram of Residuals\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Training for ≈ 10 % of data\n",
    "\n",
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
    "model2 = DenseNet(num_channels=3)\n",
    "\n",
    "model2.to(device)\n",
    "\n",
    "criterion = nn.MSELoss()  \n",
    "optimizer = torch.optim.Adam(model2.parameters(), lr=0.001)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(len(train_loader))\n",
    "print(len(train_loader) / 10)\n",
    "print(len(val_loader))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_losses = []\n",
    "val_losses = []\n",
    "batch_running_losses = []\n",
    "num_epochs = 5  \n",
    "\n",
    "for epoch in range(num_epochs):\n",
    "    start_epoch_time = time.time()\n",
    "    model2.train()\n",
    "    running_loss = 0.0     \n",
    "    \n",
    "    torch.manual_seed(2)        \n",
    "    for i, data in enumerate(train_loader, 0):      # 0 is start index, 234 is end.\n",
    "        if i >= 20:       \n",
    "            break\n",
    "        \n",
    "        batch_running_loss = 0.0\n",
    "        \n",
    "        inputs, labels = data\n",
    "        inputs, labels = inputs.to(device), labels.to(device)\n",
    "        optimizer.zero_grad()  \n",
    "        \n",
    "        outputs = model2(inputs)     \n",
    "        \n",
    "        training_loss = criterion(outputs, labels)\n",
    "        train_losses.append(training_loss.item()) \n",
    "        running_loss += training_loss.item()\n",
    "        batch_running_loss += training_loss.item()\n",
    "        \n",
    "        with open(\"train2.txt\", \"a\") as f:\n",
    "                f.write(str(training_loss.item()))\n",
    "                f.write('\\n')\n",
    "        \n",
    "        training_loss.backward()     # backward pass\n",
    "        \n",
    "        optimizer.step()             # step\n",
    "        \n",
    "        if i % 5 == 0:\n",
    "        \n",
    "            val_loss = evaluate(model2, val_loader)\n",
    "            val_losses.append(val_loss)    \n",
    "            with open(\"val2.txt\", \"a\") as f2:\n",
    "                f2.write(str(val_loss))\n",
    "                f2.write('\\n')\n",
    "                \n",
    "            model2.train()\n",
    "            \n",
    "        print('[%d, %5d] loss: %.3f' % (epoch + 1, i + 1, batch_running_loss))\n",
    "        batch_running_losses.append(batch_running_loss)\n",
    "        batch_running_loss = 0.0\n",
    "            \n",
    "        \n",
    "        avg_loss_per_epoch = running_loss / len(train_loader)\n",
    "        \n",
    "    end_epoch_time = time.time()  # End timer for epoch\n",
    "    epoch_time = end_epoch_time - start_epoch_time\n",
    "    print(f\"Epoch {epoch+1} took {epoch_time} seconds\")\n",
    "    print(f\"Avg loss for epoch {epoch+1} is  {avg_loss_per_epoch}\")\n",
    "\n",
    "\n",
    "print(\"Done!\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "losses = []\n",
    "validation = []\n",
    "\n",
    "with open('train2.txt', 'r') as f:\n",
    "    for loss in f:\n",
    "        losses.append(float(loss))\n",
    "        \n",
    "with open('val2.txt', 'r') as f2:\n",
    "    for loss in f2:\n",
    "        validation.append(float(loss))\n",
    "        \n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.plot(losses[0:20], label='Train Data, 10%')\n",
    "plt.plot(validation, label='Validation data', color='g')\n",
    "\n",
    "plt.xlabel('Batches')\n",
    "plt.ylabel('Loss')\n",
    "plt.ylim(0, 7000)\n",
    "plt.xlim(0,20)\n",
    "plt.title('Loss Trend, 10 percent of training data')\n",
    "plt.legend()\n",
    "plt.grid(True)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions, true_labels, image_paths = test_model(model2, test_loader)\n",
    "test_df = pd.DataFrame({'Image_Path': image_paths, 'Ground_Truth': true_labels, 'Inference': predictions}) # plocka ut bara komp_nivå, film och crop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "centers = test_df[test_df[\"Image_Path\"].str.contains(\"center\")]\n",
    "corners = test_df[(test_df[\"Image_Path\"].str.contains(\"bottom\") | test_df[\"Image_Path\"].str.contains(\"top\")) & (test_df[\"Image_Path\"].str.contains(\"left\") | test_df[\"Image_Path\"].str.contains(\"right\"))]\n",
    "edges = test_df[(test_df[\"Image_Path\"].str.contains(\"bottom\") | test_df[\"Image_Path\"].str.contains(\"top\") | test_df[\"Image_Path\"].str.contains(\"right\") | test_df[\"Image_Path\"].str.contains(\"left\"))] \n",
    "\n",
    "test_df['Position'] = 'Other'   # alla positioner som inte tillhör kategorierna ovan\n",
    "test_df.loc[test_df.index.isin(edges.index), 'Position'] = 'Edge'\n",
    "test_df.loc[test_df.index.isin(corners.index), 'Position'] = 'Corner'\n",
    "test_df.loc[test_df.index.isin(centers.index), 'Position'] = 'Center'\n",
    "\n",
    "test_df['Compression_Level'] = test_df['Image_Path'].apply(lambda x: x.split('_')[7].split('_frame')[0])\n",
    "test_df['Movie_Name'] = test_df['Image_Path'].apply(lambda x: x.split('_')[8].split('_frame')[0])\n",
    "\n",
    "print(test_df)\n",
    "\n",
    "test_df.to_csv('inference_10%.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_df = pd.read_csv('inference_10%.csv')\n",
    "average_ground_truth = test_df.groupby('Movie_Name')['Ground_Truth'].mean()\n",
    "\n",
    "average_vmaf_pred = test_df.groupby('Movie_Name')['Inference'].mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "corners = test_df.loc[test_df['Position'] == 'Corner']\n",
    "corners_compression_23 = corners.loc[corners['Compression_Level'] == '23']\n",
    "corners_compression_50 = corners.loc[corners['Compression_Level'] == '50']\n",
    "corners_compression_51 = corners.loc[corners['Compression_Level'] == '75'].replace('75', '51')\n",
    "\n",
    "centers = test_df.loc[test_df['Position'] == 'Center']\n",
    "centers_compression_23 = centers.loc[centers['Compression_Level'] == '23']\n",
    "centers_compression_50 = centers.loc[centers['Compression_Level'] == '50']\n",
    "centers_compression_51 = centers.loc[centers['Compression_Level'] == '75'].replace('75', '51')\n",
    "\n",
    "edges = test_df.loc[test_df['Position'] == 'Edge']\n",
    "edges_compression_23 = edges.loc[edges['Compression_Level'] == '23']\n",
    "edges_compression_50 = edges.loc[edges['Compression_Level'] == '50']\n",
    "edges_compression_51 = edges.loc[edges['Compression_Level'] == '75'].replace('75', '51')\n",
    "\n",
    "avg_vmaf_true = [sum(corners_compression_23['Ground_Truth'])/len(corners_compression_23['Ground_Truth']), sum(corners_compression_50['Ground_Truth'])/ len(corners_compression_50['Ground_Truth']), sum(corners_compression_51['Ground_Truth'])/len(corners_compression_51['Ground_Truth']), sum(centers_compression_23['Ground_Truth'])/len(centers_compression_23['Ground_Truth']), sum(centers_compression_50['Ground_Truth'])/len(centers_compression_50['Ground_Truth']), sum(centers_compression_51['Ground_Truth'])/len(centers_compression_51['Ground_Truth']), sum(edges_compression_23['Ground_Truth'])/len(edges_compression_23['Ground_Truth']), sum(edges_compression_50['Ground_Truth'])/len(edges_compression_50['Ground_Truth']), sum(edges_compression_51['Ground_Truth'])/len(edges_compression_51['Ground_Truth'])]\n",
    "avg_vmaf_pred = [sum(corners_compression_23['Inference'])/len(corners_compression_23['Inference']), sum(corners_compression_50['Inference'])/ len(corners_compression_50['Inference']), sum(corners_compression_51['Inference'])/len(corners_compression_51['Inference']), sum(centers_compression_23['Inference'])/len(centers_compression_23['Inference']), sum(centers_compression_50['Inference'])/len(centers_compression_50['Inference']), sum(centers_compression_51['Inference'])/len(centers_compression_51['Inference']), sum(edges_compression_23['Inference'])/len(edges_compression_23['Inference']), sum(edges_compression_50['Inference'])/len(edges_compression_50['Inference']), sum(edges_compression_51['Inference'])/len(edges_compression_51['Inference'])]\n",
    "\n",
    "abs_diffs_movies = []\n",
    "for movie in test_df['Movie_Name'].unique():\n",
    "    abs_diff = abs(average_ground_truth[movie] - average_vmaf_pred[movie])\n",
    "    abs_diffs_movies.append(abs_diff)\n",
    "\n",
    "r_squared = r2_score(average_ground_truth, average_vmaf_pred)\n",
    "print(\"R^2 score of model performance is\", r_squared)\n",
    "\n",
    "plt.plot(average_ground_truth, label=\"Average ground truth VMAF\")\n",
    "plt.plot(average_vmaf_pred, label=\"Average predicted VMAF\")\n",
    "plt.legend(fontsize='small')\n",
    "plt.title('Average VMAF values based on movie: true and predicted values')\n",
    "plt.xticks(rotation=45, ha='right', rotation_mode='anchor')  \n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "grouped_df = grouped_df1 = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])\n",
    "\n",
    "grouped_df1 = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Inference'].mean().reset_index()\n",
    "grouped_df2 = test_df.groupby(['Movie_Name', 'Compression_Level', 'Position'])['Ground_Truth'].mean().reset_index()\n",
    "\n",
    "mae = mean_absolute_error([grouped_df1['Inference']], [grouped_df2['Ground_Truth']], multioutput='raw_values')\n",
    "\n",
    "categories = [f\"{movie} - {compression} - {position}\" for movie, compression, position in zip(grouped_df1['Movie_Name'], grouped_df1['Compression_Level'].replace(75, 51), grouped_df1['Position'])]\n",
    "\n",
    "def colors(categories):\n",
    "    colors = []\n",
    "    for cat in categories:\n",
    "        if \"Corner\" in cat:\n",
    "            if \"23\" in cat:\n",
    "                colors.append(\"#ea9999\")\n",
    "                \n",
    "            elif \"50\" in cat:\n",
    "                colors.append(\"#e06666\")\n",
    "                \n",
    "            elif \"51\" in cat:\n",
    "                colors.append(\"#ea4c4c\")\n",
    "                \n",
    "        elif \"Center\" in cat:\n",
    "            if \"23\" in cat:\n",
    "                colors.append(\"#b6d7a8\")\n",
    "                \n",
    "            elif \"50\" in cat:\n",
    "                colors.append(\"#93c47d\")\n",
    "                \n",
    "            elif \"51\" in cat:\n",
    "                colors.append(\"#6aa84f\")\n",
    "            \n",
    "        elif \"Edge\" in cat:\n",
    "            if \"23\" in cat:\n",
    "                colors.append(\"#9fc5e8\")\n",
    "                \n",
    "            elif \"50\" in cat:\n",
    "                colors.append(\"#6fa8dc\")\n",
    "                \n",
    "            elif \"51\" in cat:\n",
    "                colors.append(\"#3d85c6\")  \n",
    "        \n",
    "    return colors\n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.bar(categories, mae, color=colors(categories))\n",
    "plt.title('Mean absolute error based on movie, compression level, and position, 10 percent trained model')\n",
    "plt.ylabel('MAE')\n",
    "plt.xticks(rotation=90)\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(\"Average error, sorted by movie, compression level and frame location is\", mae.mean())\n",
    "r_squared = r2_score(grouped_df2['Ground_Truth'], grouped_df1['Inference'])\n",
    "print(\"R^2 score of model performance, sorted by movie, compression level and frame location is\", r_squared)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define acceptable error range of 0-6.\n",
    "\n",
    "tot_errors = len(mae)\n",
    "acceptable_errors = []\n",
    "non_acceptable_errors = []\n",
    "for e in mae:\n",
    "    if e <= 6:\n",
    "        acceptable_errors.append(e)\n",
    "        \n",
    "    else:\n",
    "        non_acceptable_errors.append(e)\n",
    "\n",
    "print(tot_errors)\n",
    "      \n",
    "print(len(acceptable_errors))\n",
    "\n",
    "print(len(acceptable_errors) / tot_errors)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "indices = []\n",
    "\n",
    "for i in range(len(grouped_df1['Inference'])):\n",
    "    if mae[i] > 6:\n",
    "        indices.append(i)\n",
    "        \n",
    "categories = [f\"{movie} - {compression} - {position}\" for movie, compression, position in zip(grouped_df1['Movie_Name'][[i for i in indices]], grouped_df1['Compression_Level'].replace(75, 51)[[i for i in indices]], grouped_df1['Position'][[i for i in indices]])]\n",
    "\n",
    "\n",
    "print(indices)\n",
    "\n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.bar(categories, mae[[i for i in indices]], color=colors(categories))\n",
    "plt.title('Mean absolute error, outside allowed error range (>6)')\n",
    "plt.ylabel('MAE')\n",
    "plt.xticks(rotation=90)\n",
    "plt.tight_layout()\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "folder_dir = \"images_train\"\n",
    "ims = []\n",
    "for image in os.listdir(folder_dir):\n",
    "    ims.append(os.path.join('images_train/', image))\n",
    "    \n",
    "print(plt.imread(ims[0]).shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Future DenseNet implementation, test.\n",
    "class DenseNet(nn.Module):\n",
    "   def __init__(self, num_channels=3):     \n",
    "       super().__init__()\n",
    "       preloaded = models.densenet121(pretrained=True)\n",
    "       self.features = preloaded.features\n",
    "       self.features.conv0 = nn.Conv2d(num_channels, 64, 7, 2, 3)   \n",
    "       nn.Linear(50176, 1024),            \n",
    "       nn.ReLU(inplace=True),\n",
    "      \n",
    "       for param in self.features.parameters():  # Freeze pretrained model features\n",
    "           param.requires_grad = False\n",
    "          \n",
    "   \n",
    "   def forward(self, x):     \n",
    "       features = self.features(x)       # hidden layers, feature maps\n",
    "       out = F.relu(features, inplace=True)     # activation function\n",
    "       \n",
    "       return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "annotations = pd.read_csv(\"vmaf_on_full_frames_train_ALL.csv\")\n",
    "img_path = annotations.iloc[1, 1]\n",
    "movie_frame_info = img_path.split(\"_\")[5:]\n",
    "patch_list = []\n",
    "patches = []\n",
    "\n",
    "image_groups = {}\n",
    "\n",
    "for image_name in os.listdir(\"images_train_comp_crop_png\"):\n",
    "    image_path = os.path.join(\"images_train_comp_crop_png\", image_name)\n",
    "    movie, frame = image_name.split(\"_\")[4], image_name.split(\"_\")[5]\n",
    "    \n",
    "    key = (movie, frame)\n",
    "    if key not in image_groups:\n",
    "        image_groups[key] = []\n",
    "    image_groups[key].append(image_path)\n",
    "\n",
    "patches = list(image_groups.values())\n",
    "\n",
    "# print(patches[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "f = os.listdir(\"images_train_comp_crop_png\")[0]\n",
    "f_split = f.split(\"_\")\n",
    "print(f)\n",
    "print(f_split[0:5])     \n",
    "print(f_split[5])\n",
    "root_dir = \"images_train_comp_crop_png\"\n",
    "plt.imshow(mpimg.imread(os.path.join(root_dir, str(f))))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class ImagesAndLabelsDataset(Dataset):\n",
    "    def __init__(self, csv_file, root_dir, patches, transform=None):    # num_patches=(1080//224, 1920//224)\n",
    "        self.annotations = pd.read_csv(csv_file)  # labels\n",
    "        self.root_dir = root_dir  # image folder\n",
    "        self.patches = patches\n",
    "        self.transform = transform\n",
    "      \n",
    "    def __len__(self):\n",
    "        return len(self.annotations)\n",
    "  \n",
    "    def __getitem__(self, index):\n",
    "        for im in patches: \n",
    "            \n",
    "            self.dense_layers = nn.ModuleList([nn.Sequential(\n",
    "            nn.Linear(50176, 1024),            \n",
    "            nn.ReLU(inplace=True),\n",
    "            nn.Linear(1024, 1024),\n",
    "            nn.ReLU(inplace=True),\n",
    "            nn.Linear(1024, 1024),\n",
    "            nn.ReLU(inplace=True),\n",
    "            nn.Linear(1024, 1024),\n",
    "            nn.ReLU(inplace=True),\n",
    "            nn.Linear(1024, 1024),\n",
    "            nn.ReLU(inplace=True),\n",
    "            nn.Linear(1024, 1)\n",
    "            )])\n",
    "      \n",
    "            # unfreeze the last Dense layers:\n",
    "            for dense in self.dense_layers:\n",
    "                for param in dense.parameters():\n",
    "                    param.requires_grad = True\n",
    "                    \n",
    "                    \n",
    "            img_path = os.listdir(self.root_dir)[index]\n",
    "            # List with compression info, movie name and frame number.\n",
    "            movie_frame_info = img_path.split(\"_\")[0:5]  \n",
    "            \n",
    "            try:\n",
    "                image = mpimg.imread(os.path.join(root_dir, str(img_path)))\n",
    "                \n",
    "            except Exception as e:\n",
    "                print(f\"Error reading image at index {index}: {e}\")\n",
    "                return None\n",
    "            \n",
    "            norm = transforms.Compose([transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])   # From https://pytorch.org/hub/pytorch_vision_densenet/\n",
    "        \n",
    "            if self.transform:\n",
    "                image = self.transform(im)\n",
    "                norm(image)\n",
    "        \n",
    "        # VMAF score based on the full frame\n",
    "        full_frame_vmaf = self.annotations.iloc[index, 0]\n",
    "            \n",
    "        patches_tensor = torch.stack(patches)  \n",
    "        full_frame_vmaf_tensor = torch.tensor(full_frame_vmaf, dtype=torch.float32)\n",
    "        movie_frame_info_tensor = torch.tensor(movie_frame_info.encode(), dtype=torch.uint8)\n",
    "          \n",
    "        return (patches_tensor, full_frame_vmaf_tensor, movie_frame_info)    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = ImagesAndLabelsDataset(csv_file = \"vmaf_on_full_frames_train.csv\", root_dir = \"images_train_comp_crop_png\", patches = patches, transform = transforms.ToTensor())\n",
    "train_set, val_set = torch.utils.data.random_split(data, [0.9, 0.1], generator=torch.Generator().manual_seed(1))    # 10% validation\n",
    "train_loader = DataLoader(dataset=train_set, shuffle=True, batch_size=100)   \n",
    "val_loader = DataLoader(dataset=val_set, shuffle=True, batch_size=100) \n",
    "\n",
    "\n",
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
    "model2 = DenseNet(num_channels=3)       # (num_patch, 3, 224, 224)\n",
    "\n",
    "model2.to(device)\n",
    "\n",
    "criterion = nn.MSELoss() \n",
    "optimizer = torch.optim.Adam(model2.parameters(), lr=0.001) \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_losses = []\n",
    "val_losses = []\n",
    "batch_running_losses = []\n",
    "num_epochs = 5  \n",
    "\n",
    "for epoch in range(num_epochs):\n",
    "    start_epoch_time = time.time()\n",
    "    model.train() \n",
    "    \n",
    "    torch.manual_seed(2)        \n",
    "    for i, data in enumerate(train_loader, 0):      \n",
    "        batch_running_loss = 0.0\n",
    "        \n",
    "        inputs, labels, info = data\n",
    "        inputs, labels = inputs, labels\n",
    "        optimizer.zero_grad()  \n",
    "        \n",
    "        outputs = model(inputs)     \n",
    "        \n",
    "        training_loss = criterion(outputs, labels)\n",
    "        train_losses.append(training_loss.item()) \n",
    "        \n",
    "        with open(\"train3.txt\", \"a\") as f:\n",
    "                f.write(str(training_loss.item()))\n",
    "                f.write('\\n')\n",
    "        \n",
    "        training_loss.backward()     # backward pass\n",
    "        \n",
    "        optimizer.step()             # step\n",
    "        \n",
    "              \n",
    "        val_loss = evaluate(model, val_loader)\n",
    "        val_losses.append(val_loss)    \n",
    "        with open(\"val3.txt\", \"a\") as f2:\n",
    "            f2.write(str(val_loss))\n",
    "            f2.write('\\n')\n",
    "                \n",
    "        model.train()\n",
    "            \n",
    "        print('[%d, %5d] loss: %.3f' % (epoch + 1, i + 1, batch_running_loss / 25))\n",
    "        batch_running_losses.append(batch_running_loss / 25)\n",
    "        batch_running_loss = 0.0\n",
    "            \n",
    "        \n",
    "        avg_loss_per_epoch = running_loss / len(train_loader)\n",
    "        \n",
    "    end_epoch_time = time.time()  # End timer for epoch\n",
    "    epoch_time = end_epoch_time - start_epoch_time\n",
    "    print(f\"Epoch {epoch+1} took {epoch_time} seconds\")\n",
    "    print(f\"Avg loss for epoch {epoch+1} is  {avg_loss_per_epoch}\")\n",
    "\n",
    "\n",
    "print(\"Done!\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_losses = []\n",
    "val_losses = []\n",
    "num_epochs = 5  \n",
    "\n",
    "for epoch in range(num_epochs):\n",
    "    start_epoch_time = time.time()\n",
    "    model.train() \n",
    "    \n",
    "    torch.manual_seed(2)        \n",
    "    for i, data in enumerate(train_loader, 0):  \n",
    "        inputs, labels, info = data\n",
    "        inputs, labels = inputs.to(device), labels.to(device)  \n",
    "        optimizer.zero_grad()  \n",
    "        \n",
    "        outputs = model(inputs)\n",
    "        \n",
    "        training_loss = criterion(outputs, labels)\n",
    "        train_losses.append(training_loss.item()) \n",
    "        \n",
    "        training_loss.backward()\n",
    "        optimizer.step()\n",
    "            \n",
    "        if i % 25 == 24:  # Print every 25 mini-batches\n",
    "            print('[%d, %5d] loss: %.3f' % (epoch + 1, i + 1, running_loss / 25))\n",
    "        \n",
    "    end_epoch_time = time.time()\n",
    "    epoch_time = end_epoch_time - start_epoch_time\n",
    "    print(f\"Epoch {epoch+1} took {epoch_time} seconds\")\n",
    "    \n",
    "    # Evaluate on validation set after each epoch\n",
    "    val_loss = evaluate(model, val_loader)\n",
    "    val_losses.append(val_loss)\n",
    "    print(f\"Avg validation loss for epoch {epoch+1} is {val_loss}\")\n",
    "\n",
    "print(\"Training done!\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def frame_locations(df): \n",
    "    corners = test_df.loc[test_df['Position'] == 'Corner']\n",
    "    corners_compression_23 = corners.loc[corners['Compression_Level'] == '23']\n",
    "    corners_compression_50 = corners.loc[corners['Compression_Level'] == '50']\n",
    "    corners_compression_75 = corners.loc[corners['Compression_Level'] == '75']\n",
    "\n",
    "    centers = test_df.loc[test_df['Position'] == 'Center']\n",
    "    centers_compression_23 = centers.loc[centers['Compression_Level'] == '23']\n",
    "    centers_compression_50 = centers.loc[centers['Compression_Level'] == '50']\n",
    "    centers_compression_75 = centers.loc[centers['Compression_Level'] == '75']\n",
    "\n",
    "    edges = test_df.loc[test_df['Position'] == 'Edge']\n",
    "    edges_compression_23 = edges.loc[edges['Compression_Level'] == '23']\n",
    "    edges_compression_50 = edges.loc[edges['Compression_Level'] == '50']\n",
    "    edges_compression_75 = edges.loc[edges['Compression_Level'] == '75']\n",
    "    \n",
    "    return [corners_compression_23, corners_compression_50, corners_compression_75, centers_compression_23, centers_compression_50, centers_compression_75, edges_compression_23, edges_compression_50, edges_compression_75]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE loss based on movie name is 9.186635124423029\n",
      "R^2 score of model performance, sorted by movie is 0.8157819526742741\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "test_df = pd.read_csv('inference.csv')\n",
    "average_ground_truth = test_df.groupby('Movie_Name')['Ground_Truth'].mean()\n",
    "average_vmaf_pred = test_df.groupby('Movie_Name')['Inference'].mean()\n",
    "ground_truths = test_df[\"Ground_Truth\"]\n",
    "predictions = test_df[\"Inference\"]\n",
    "abs_diffs_movies = []\n",
    "for movie in test_df['Movie_Name'].unique():\n",
    "    abs_diff = abs(average_ground_truth[movie] - average_vmaf_pred[movie])\n",
    "    abs_diffs_movies.append(abs_diff)\n",
    "    \n",
    "print(\"MAE loss based on movie name is\", sum(abs_diffs_movies)/len(abs_diffs_movies))\n",
    "\n",
    "r_squared = r2_score(average_ground_truth, average_vmaf_pred)\n",
    "print(\"R^2 score of model performance, sorted by movie is\", r_squared)\n",
    "\n",
    "plt.plot(average_ground_truth, label=\"Average ground truth VMAF\")\n",
    "plt.plot(average_vmaf_pred, label=\"Average predicted VMAF\")\n",
    "plt.xticks(rotation=45, ha='right', rotation_mode='anchor')  \n",
    "plt.legend(fontsize='small')\n",
    "plt.title('Average VMAF values based on movie: true and predicted values')\n",
    "plt.grid(alpha=0.2)\n",
    "plt.show()\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['agent-cody-banks.mp4', 'apocalypse-now-final-cut.mp4', 'carrie-2013.mp4', 'domestic-disturbance-2001.mp4', 'fading-gigolo.mp4', 'havenhurst.mp4', 'karsten-och-petra-pa-safari.mp4', 'karsten-og-petras-vidunderlige-jul.mp4', 'little-vampire-2020.mp4', 'maria-wern-pojke-forsvunnen.mp4', 'maria-wern-stum-sitter-guden.mp4', 'play.mp4', 'robocop-2.mp4', 'rudderless.mp4', 'the-apartment.mp4', 'the-mauritanian.mp4', 'the-sun-is-also-a-star.mp4']\n"
     ]
    }
   ],
   "source": [
    "movies = [mov for mov in test_df['Movie_Name'].unique()]\n",
    "print(movies)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x = np.arange(len(movies))\n",
    "bar_width = 0.25\n",
    "\n",
    "plt.bar(x - bar_width/2, average_ground_truth, width=bar_width, label=\"Average ground truth VMAF\", alpha=0.95)\n",
    "plt.bar(x + bar_width/2, average_vmaf_pred, width=bar_width, label=\"Average predicted VMAF\", alpha=0.95)\n",
    "\n",
    "plt.xticks(x, movies, rotation=50, ha='right', rotation_mode='anchor')  \n",
    "plt.legend(fontsize='small', loc='lower right')\n",
    "plt.title('Average VMAF values based on movie: true and predicted values')\n",
    "plt.grid(alpha=0.5)\n",
    "plt.tight_layout()  \n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.74405189568246e-90\n",
      "Reject the null hypothesis: There is a difference between the population means.\n"
     ]
    }
   ],
   "source": [
    "# Statistical significance test and P-value:\n",
    "from scipy.stats import ttest_rel\n",
    "\n",
    "df = pd.read_csv('inference.csv')\n",
    "\n",
    "pred = df['Inference']    # sample population\n",
    "gt = df['Ground_Truth']   # true population\n",
    "\n",
    "pred_mean = np.mean(pred)\n",
    "pred_std= np.std(pred)\n",
    "true_mean = np.mean(gt)\n",
    "true_std = np.std(gt)\n",
    "\n",
    "t_test = ttest_rel(gt, pred)\n",
    "print(t_test.pvalue)\n",
    "\n",
    "significance_level = 0.05\n",
    "if t_test.pvalue < significance_level:\n",
    "    print(\"Reject the null hypothesis: There is a difference between the population means.\")\n",
    "else:\n",
    "    print(\"Fail to reject the null hypothesis: There is no significant difference between the population means.\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.74405189568246e-90\n",
      "Reject the null hypothesis: There is a difference between the population means.\n"
     ]
    }
   ],
   "source": [
    "# Statistical significance test and P-value:\n",
    "# Dependent samples\n",
    "from scipy.stats import ttest_rel\n",
    "\n",
    "df = pd.read_csv('inference.csv')\n",
    "\n",
    "pred = df['Inference']    # sample population\n",
    "gt = df['Ground_Truth']   # true population\n",
    "\n",
    "t_test = ttest_rel(pred, gt)\n",
    "print(t_test.pvalue)\n",
    "\n",
    "significance_level = 0.01\n",
    "if t_test.pvalue < significance_level:\n",
    "    print(\"Reject the null hypothesis: There is a difference between the population means.\")\n",
    "else:\n",
    "    print(\"Fail to reject the null hypothesis: There is no significant difference between the population means.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-13.893523172889031\n"
     ]
    }
   ],
   "source": [
    "import math\n",
    "\n",
    "n = 5920\n",
    "\n",
    "# Beräkna t-värde\n",
    "t = (true_mean - pred_mean) / math.sqrt(((true_std**2)/n) + ((pred_std**2)/n))\n",
    "\n",
    "print(t)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-804.0013538399309\n"
     ]
    }
   ],
   "source": [
    "# Control of t-test\n",
    "import math\n",
    "\n",
    "t = (true_mean - pred_mean) / math.sqrt(((true_std)/5920)**2 + ((pred_std)/5920)**2)\n",
    "print(t)\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.10.5 64-bit",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.5"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "aee8b7b246df8f9039afb4144a1f6fd8d2ca17a180786b69acc140d282b71a49"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}