visualization.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "os.chdir('/home/jckim/workspace/ASFormer')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = 'gtea' \n",
    "mode='rand'\n",
    "train_mode=''\n",
    "\n",
    "result_dir='results'\n",
    "testing='y'\n",
    "ndir='/data/jckim/data/'\n",
    "\n",
    "if train_mode != '':\n",
    "   result_dir=f'results_{train_mode}'\n",
    "if mode != '':\n",
    "  testing='n'\n",
    "  result_dir += f'_{mode}'\n",
    "  tdir=f'/data/jckim/data_{mode}/' # /data/jckim/data_mode/\n",
    "\n",
    "if dataset == '50salads':\n",
    "    split = 5\n",
    "else:\n",
    "    split = 4"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def read_file(path):\n",
    "    with open(path, 'r') as f:\n",
    "        content = f.read()\n",
    "        f.close()\n",
    "    return content"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "if dataset=='gtea':\n",
    "  # lists={'take': 'background', \"close\": \"put\", \"open\": \"background\"}\n",
    "  lists = {\"close\": \"put\"}\n",
    "# elif dataset=='50salads':\n",
    "#   lists={'cut_tomato': 'place_tomato_into_bowl', \"peel_cucumber\": \"cut_cucumber\", \"cut_cucumber\" :'place_cucumber_into_bowl', \\\n",
    "# 'cut_lettuce': 'place_lettuce_into_bowl'}\n",
    "# else:\n",
    "#   lists={\"pour_dough2pan\": \"fry_pancake\", \"put_egg2plate\": \"SIL\", \"cut_orange\": \"squeeze_orange\", \"put_bunTogether\": \"SIL\"}\n",
    "elif dataset=='breakfast':  \n",
    "  lists = {\"pour_dough2pan\": \"fry_pancake\"}\n",
    "  \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "if dataset=='gtea':\n",
    "  label='close'\n",
    "elif dataset=='50salads':\n",
    "  label='cut_tomato'\n",
    "else:\n",
    "  label=\"pour_dough2pan\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "correct_gt = 0\n",
    "total_gt = 0\n",
    "a=0\n",
    "b=0\n",
    "ground_truth_counts = {}\n",
    "prediction_counts = {}\n",
    "\n",
    "for s in range(1, split+1):\n",
    "  recog_path = \"./{}/\".format(result_dir)+dataset+\"/split_{}\".format(s)+\"/\"\n",
    "  file_list = ndir +dataset+\"/splits/test.split{}\".format(s)+\".bundle\"\n",
    "  if testing == 'y':\n",
    "    ground_truth_path = tdir + dataset + \"/groundTruth/\"\n",
    "    mapping_file = ndir + dataset + \"/mapping.txt\"\n",
    "    list_of_videos = read_file(file_list).split('\\n')[:-1]\n",
    "  else:\n",
    "    ground_truth_path = ndir + dataset + \"/groundTruth/\"\n",
    "    mapping_file = ndir + dataset + \"/mapping.txt\"\n",
    "    list_of_videos = read_file(file_list).split('\\n')[:-1]\n",
    "\n",
    "  for vid in list_of_videos:\n",
    "    gt_file = ground_truth_path + vid\n",
    "    gt_content = read_file(gt_file).split('\\n')[0:-1]\n",
    "    new_list = []\n",
    "    last_element = None\n",
    "    for element in gt_content:\n",
    "      if element != last_element:\n",
    "        new_list.append(element)\n",
    "        last_element = element\n",
    "    gt_content = new_list\n",
    "    \n",
    "    if testing == 'y':\n",
    "      recog_file = str(\"/\".join(recog_path.split('/')[:-2])) + '/' + str(recog_path.split('/')[-2] + '/' + vid.split('.')[0])\n",
    "    else:\n",
    "      recog_file = recog_path  + vid.split('.')[0]\n",
    "    recog_content = read_file(recog_file).split('\\n')[1].split()\n",
    "    new_list = []\n",
    "    last_element = None\n",
    "    for element in recog_content:\n",
    "      if element != last_element:\n",
    "        new_list.append(element)\n",
    "        last_element = element\n",
    "    recog_content = new_list\n",
    "\n",
    "    for i in range(len(gt_content)-1):\n",
    "      if gt_content[i] in lists.keys():\n",
    "        total_gt += 1\n",
    "        if gt_content[i+1] == lists[gt_content[i]]:\n",
    "          correct_gt += 1\n",
    "      if gt_content[i] == label:\n",
    "        if gt_content[i+1] in ground_truth_counts.keys():\n",
    "          ground_truth_counts[gt_content[i+1]] += 1\n",
    "        else:\n",
    "          ground_truth_counts[gt_content[i+1]] = 1\n",
    "    \n",
    "    for i in range(len(recog_content)-1):\n",
    "      if recog_content[i] in lists.keys():\n",
    "        a += 1\n",
    "        if recog_content[i+1] == lists[recog_content[i]]:\n",
    "          b += 1\n",
    "      if recog_content[i] == label:\n",
    "        if recog_content[i+1] in prediction_counts.keys():\n",
    "          prediction_counts[recog_content[i+1]] += 1\n",
    "        else:\n",
    "          prediction_counts[recog_content[i+1]] = 1\n",
    "\n",
    "        \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "real_counts={}\n",
    "\n",
    "for s in range(1, split+1):\n",
    "  recog_path = \"./{}/\".format(result_dir)+dataset+\"/split_{}\".format(s)+\"/\"\n",
    "  file_list = ndir +dataset+\"/splits/test.split{}\".format(s)+\".bundle\"\n",
    "  ground_truth_path = ndir + dataset + \"/groundTruth/\"\n",
    "  mapping_file = ndir + dataset + \"/mapping.txt\"\n",
    "  list_of_videos = read_file(file_list).split('\\n')[:-1]\n",
    "\n",
    "  for vid in list_of_videos:\n",
    "    gt_file = ground_truth_path + vid\n",
    "    gt_content = read_file(gt_file).split('\\n')[0:-1]\n",
    "\n",
    "    new_list = []\n",
    "    last_element = None\n",
    "    for element in gt_content:\n",
    "      if element != last_element:\n",
    "        new_list.append(element)\n",
    "        last_element = element\n",
    "    gt_content = new_list\n",
    "\n",
    "    \n",
    "    if testing == 'y':\n",
    "      recog_file = str(\"/\".join(recog_path.split('/')[:-2])) + '/' + str(recog_path.split('/')[-2] + '/' + vid.split('.')[0])\n",
    "    else:\n",
    "      recog_file = recog_path  + vid.split('.')[0]\n",
    "    recog_content = read_file(recog_file).split('\\n')[1].split()\n",
    "\n",
    "    new_list = []\n",
    "    last_element = None\n",
    "    for element in recog_content:\n",
    "      if element != last_element:\n",
    "        new_list.append(element)\n",
    "        last_element = element\n",
    "    recog_content = new_list\n",
    "\n",
    "    for i in range(len(gt_content)-1):\n",
    "      if gt_content[i] == label:\n",
    "        if gt_content[i+1] in real_counts.keys():\n",
    "          real_counts[gt_content[i+1]] += 1\n",
    "        else:\n",
    "          real_counts[gt_content[i+1]] = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'put': 64, 'background': 3}\n",
      "{'put': 54, 'background': 7, 'open': 3, 'stir': 1}\n",
      "{'put': 64, 'background': 3}\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[Text(0, 0, 'background'),\n",
       " Text(1, 0, 'open'),\n",
       " Text(2, 0, 'put'),\n",
       " Text(3, 0, 'stir')]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1500x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(ground_truth_counts)\n",
    "print(prediction_counts)\n",
    "print(real_counts)\n",
    "\n",
    "# Get unique keys from both dictionaries\n",
    "keys = set(list(real_counts.keys()) + list(ground_truth_counts.keys()) + list(prediction_counts.keys()))\n",
    "\n",
    "# Extract values for each key from both dictionaries\n",
    "real_vals = [real_counts.get(key, 0) for key in keys]\n",
    "groundtruth_vals = [ground_truth_counts.get(key, 0) for key in keys]\n",
    "prediction_vals = [prediction_counts.get(key, 0) for key in keys]\n",
    "\n",
    "# Set up the plot\n",
    "fig, ax = plt.subplots(figsize=(15, 5))\n",
    "plt.rc('font', size=20) \n",
    "index = range(len(keys))\n",
    "bar_width = 0.25\n",
    "opacity = 0.8\n",
    "\n",
    "# Create bars for ground truth\n",
    "\n",
    "groundtruth_bars = ax.bar(index, groundtruth_vals, bar_width,\n",
    "                 alpha=opacity,\n",
    "                 color='b',\n",
    "                 label='Ground Truth')\n",
    "\n",
    "real_bars = ax.bar([i - bar_width for i in index], real_vals, bar_width,\n",
    "                  alpha=opacity,\n",
    "                  color='r',\n",
    "                  label='Original')\n",
    "\n",
    "# Create bars for prediction\n",
    "prediction_bars = ax.bar([i + bar_width for i in index], prediction_vals, bar_width,\n",
    "                 alpha=opacity,\n",
    "                 color='g',\n",
    "                 label='Prediction')\n",
    "\n",
    "# Add labels, title, and legend\n",
    "ax.set_xlabel('Action Label')\n",
    "ax.set_ylabel('Count')\n",
    "ax.set_title('Model: ASFormer, Dataset: GTEA')\n",
    "ax.set_xticks([i for i in index])\n",
    "ax.set_xticklabels(keys)\n",
    "# ax.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9552238805970149\n",
      "0.8307692307692308\n",
      "67 64\n"
     ]
    }
   ],
   "source": [
    "print(correct_gt/total_gt)\n",
    "print(b/a)\n",
    "print(total_gt, correct_gt)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.10.8"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "8c9208282c268ab09878474c077c8c9a0aa3e8118fc3d237ff48ed7373f673ae"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}