performance_metrics.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.optim as optim\n",
    "import torch.utils.data as data\n",
    "from torch.utils.data import DataLoader, sampler, TensorDataset\n",
    "from torch.utils.data import sampler\n",
    "\n",
    "import torch.nn.functional as F\n",
    "\n",
    "import torchvision.datasets as dset\n",
    "import torchvision.transforms as T\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import numpy as np\n",
    "import pickle\n",
    "import os\n",
    "from PIL import Image\n",
    "\n",
    "from scipy.interpolate import interp2d\n",
    "from scipy.interpolate import RectBivariateSpline\n",
    "from scipy.ndimage.filters import gaussian_filter\n",
    "\n",
    "import h5py\n",
    "\n",
    "from time import time\n",
    "\n",
    "from classes import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "using device: cuda\n"
     ]
    }
   ],
   "source": [
    "# From Assignment 2 PyTorch Notebook\n",
    "USE_GPU = True\n",
    "\n",
    "dtype = torch.float32 # we will be using float throughout this tutorial\n",
    "\n",
    "if USE_GPU and torch.cuda.is_available():\n",
    "    device = torch.device('cuda')\n",
    "else:\n",
    "    device = torch.device('cpu')\n",
    "\n",
    "# Constant to control how frequently we print train loss\n",
    "print_every = 1\n",
    "\n",
    "print('using device:', device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_directory = '.'\n",
    "\n",
    "##########################################################################################\n",
    "#                                  BATCH_SIZE PARAMETER\n",
    "BATCH_SIZE = 1000\n",
    "##########################################################################################\n",
    "\n",
    "\n",
    "# sr_train = SR_Dataset(data_directory, train=TRAIN_MODE)\n",
    "# loader_train = DataLoader(sr_train, batch_size=BATCH_SIZE, \n",
    "#                           sampler=sampler.SubsetRandomSampler(range(TRAIN_FILES*YEARS_PER_FILE*IMGS_PER_YEAR)),\n",
    "#                          num_workers = 4)\n",
    "\n",
    "\n",
    "sr_val = SR_Dataset(data_directory, train=VAL_MODE)\n",
    "loader_val = DataLoader(sr_val, batch_size=BATCH_SIZE, \n",
    "                          sampler=sampler.SubsetRandomSampler(range(VAL_FILES*YEARS_PER_FILE*IMGS_PER_YEAR)),\n",
    "                         num_workers = 4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "# input shape: (days, height, width). calculates rmse of each pixel.\n",
    "# or input shape: (N, C, H, W), same thing.\n",
    "def rmse_pixel(gen, target):\n",
    "    return np.sqrt(np.mean(np.square(gen - target), axis = 0))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "# y, y_pred: (N, C, H, W)\n",
    "def get_extremes(y, y_pred, quantiles, metric_fn):\n",
    "    output = np.zeros((2, len(quantiles), 2)) # channel, quantile, [mean/variance]\n",
    "    for q_index, q in enumerate(quantiles):\n",
    "        y_precip = y[:, 0, :, :]\n",
    "        y_pred_precip = y_pred[:, 0, :, :]\n",
    "        y_temp = y[:, 1, :, :]\n",
    "        y_pred_temp = y_pred[:, 1, :, :]\n",
    "        \n",
    "        H = y.shape[2]\n",
    "        W = y.shape[3]\n",
    "        # [precip|temp]_indices are (num_values*H*W, 3) arrays, where num_values is the number of values\n",
    "        # above this percentile, and the 3 columns correspodn to the three coordinates (see the for loops)\n",
    "        # below\n",
    "        precip_indices = np.array(np.where(y_precip > np.percentile(a = y_precip, q = q, axis = 0))).T\n",
    "        temp_indices = np.array(np.where(y_temp > np.percentile(a = y_temp, q = q, axis = 0))).T\n",
    "        num_values = int(precip_indices.shape[0] / (H * W)) + 1\n",
    "        y_precip_extreme = np.zeros((num_values, H, W))\n",
    "        y_pred_precip_extreme = np.zeros((num_values, H, W))\n",
    "        y_temp_extreme = np.zeros((num_values, H, W))\n",
    "        y_pred_temp_extreme = np.zeros((num_values, H, W))\n",
    "#         if num_values <= 2: continue\n",
    "        # coordinates aren't in order of pixel, so keep track of how many values we've layered\n",
    "        # for each pixel so far\n",
    "        counters = np.zeros((H, W), dtype = int)\n",
    "        for i, j, k in precip_indices:\n",
    "            y_precip_extreme[counters[j, k], j, k] = y_precip[i, j, k]\n",
    "            y_pred_precip_extreme[counters[j, k], j, k] = y_pred_precip[i, j, k]\n",
    "            counters[j, k] += 1\n",
    "        counters = np.zeros((H, W), dtype = int)\n",
    "        for i, j, k in temp_indices:\n",
    "            y_temp_extreme[counters[j, k], j, k] = y_temp[i, j, k]\n",
    "            y_pred_temp_extreme[counters[j, k], j, k] = y_pred_temp[i, j, k]\n",
    "            counters[j, k] += 1\n",
    "\n",
    "        metric_precip = metric_fn(y_precip_extreme, y_pred_precip_extreme)\n",
    "        metric_temp = metric_fn(y_temp_extreme, y_pred_temp_extreme)\n",
    "        output[0, q_index, 0] = np.mean(metric_precip)\n",
    "        output[0, q_index, 1] = np.var(metric_precip)\n",
    "        output[1, q_index, 0] = np.mean(metric_temp)\n",
    "        output[1, q_index, 1] = np.var(metric_temp)\n",
    "    return output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [],
   "source": [
    "# G is generic for any image-generating model; could be SRCNN too\n",
    "def get_metrics(loader, G):\n",
    "    G = G.to(device=device)\n",
    "    G.eval() # set model to evaluation mode\n",
    "    \n",
    "    # initialize metrics\n",
    "    final_rmse = None\n",
    "    extreme_quantiles = [i / 10 for i in range(900, 1000, 5)]\n",
    "    extreme_metrics = [] # list will be populated with metrics from each batch\n",
    "    count = 0\n",
    "    \n",
    "    # calculate all metrics in one pass through val set\n",
    "    with torch.no_grad():\n",
    "        count = 0\n",
    "        for x, y in loader:\n",
    "            x = x.to(device=device, dtype=dtype) # move to device, e.g. GPU\n",
    "            y = y.to(device=device, dtype=dtype)\n",
    "            y_pred = G(x)\n",
    "            y = y.cpu().numpy()\n",
    "            y_pred = y_pred.cpu().numpy()\n",
    "            if final_rmse is not None:\n",
    "                final_rmse += rmse_pixel(y, y_pred)\n",
    "            else:\n",
    "                final_rmse = rmse_pixel(y, y_pred)\n",
    "            extreme_metrics.append(get_extremes(y, y_pred, extreme_quantiles, rmse_pixel))\n",
    "            count += 1\n",
    "            print('Done with batch %d.' % count)\n",
    "        final_rmse /= count\n",
    "        # num_batches, channel, quantile, [mean|variance]\n",
    "        extreme_metrics = np.array(extreme_metrics)\n",
    "        # channel, quantile, [mean|variance]\n",
    "        extreme_metrics = np.mean(extreme_metrics, axis = 0)\n",
    "    return final_rmse, extreme_metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Done with batch 1.\n",
      "Done with batch 2.\n",
      "Done with batch 3.\n",
      "Done with batch 4.\n",
      "Done with batch 5.\n",
      "Done with batch 6.\n",
      "Done with batch 7.\n",
      "Done with batch 8.\n"
     ]
    }
   ],
   "source": [
    "G = Generator(num_channels=3)\n",
    "\n",
    "# Load the states from training on full dataset\n",
    "G.load_state_dict(torch.load('SRGAN_G_06_05_best_precip.pt'))\n",
    "\n",
    "final_rmse, extreme_metrics = get_metrics(loader_val, G)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f8919617b00>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "extreme_quantiles = [i / 10 for i in range(900, 1000, 5)]\n",
    "# quantile, [mean|variance]\n",
    "extreme_metrics_precip = extreme_metrics[0, :, :]\n",
    "plt.fill_between(\n",
    "    extreme_quantiles, \n",
    "    extreme_metrics_precip[:, 0] - np.sqrt(extreme_metrics_precip[:, 1]),\n",
    "    extreme_metrics_precip[:, 0] + np.sqrt(extreme_metrics_precip[:, 1]),\n",
    ")\n",
    "plt.plot(extreme_quantiles, extreme_metrics_precip[:, 0], 'k')\n",
    "plt.xlabel('Quantile')\n",
    "plt.ylabel('RMSE')\n",
    "plt.title('GAN RMSE for Extreme Values')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f891981b588>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(final_rmse[0, :, :], cmap = 'hot')\n",
    "#plt.title('RMSE Map of Precipitation for GAN')\n",
    "plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f89198cf9b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(final_rmse[1, :, :], cmap = 'hot')\n",
    "#plt.title('RMSE Map of Temperature for GAN')\n",
    "plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import csv\n",
    "\n",
    "with open('extreme_analysis_SRGAN.csv', 'w') as extreme_csv:\n",
    "    writer = csv.writer(extreme_csv)\n",
    "    writer.writerow(['Quantile', 'PrecipMean', 'PrecipVar', 'TempMean', 'TempVar'])\n",
    "    for i in range(len(extreme_quantiles)):\n",
    "        writer.writerow([extreme_quantiles[i]] + list(extreme_metrics[0,i,:]) + list(extreme_metrics[1,i,:]))"
   ]
  }
 ],
 "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}