OCCuntrained_test.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "3890205e-7b52-4a1c-9884-043330334306",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training on device cpu.\n",
      "<class 'torch.Tensor'>\n",
      "Concatenated TESTING images: 435 and labels pair: 435 :\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "##############################\n",
    "#Training loop for NNs       #\n",
    "#Maintainer: Christopher Chan#\n",
    "#Version: 0.3.2              #\n",
    "#Date: 2022-05-29            #\n",
    "##############################\n",
    "\n",
    "import os\n",
    "import sys\n",
    "import torch\n",
    "import pathlib\n",
    "import time\n",
    "import re\n",
    "import PIL\n",
    "import random\n",
    "#import ray\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import torch.nn as nn\n",
    "import segmentation_models_pytorch as smp\n",
    "#from ray import tune\n",
    "#from ray.tune.suggest.bohb import TuneBOHB\n",
    "#from ray.tune.schedulers import HyperBandForBOHB\n",
    "from scipy import ndimage\n",
    "from PIL import Image\n",
    "from torch import optim\n",
    "from tqdm import tqdm\n",
    "from torchvision import transforms, datasets\n",
    "from torchvision.utils import save_image\n",
    "from torch.utils.data import DataLoader, random_split, ChainDataset, ConcatDataset\n",
    "from torch.utils.tensorboard import SummaryWriter\n",
    "from Networks import Five_UNet, Four_UNet\n",
    "from dataloader import BuildingDataset\n",
    "%matplotlib inline\n",
    "\n",
    "device = (torch.device(\"cuda\") if torch.cuda.is_available()\n",
    "          else torch.device(\"cpu\"))\n",
    "\n",
    "print(f\"Training on device {device}.\")\n",
    "\n",
    "data_path = os.path.abspath(\"/home/chris/Dropbox/HOTOSM/SAMPLE\")\n",
    "\n",
    "TestIMG_ls = []\n",
    "TestLBL_ls = []\n",
    "\n",
    "for root, dirs, filename in os.walk(os.path.join(data_path, \"Test_15cm\", \"IMG\")):\n",
    "    for i in filename:\n",
    "        #if i.startswith(\"KBY15_IMG\"):\n",
    "            TestIMG_ls.append(root + \"/\" + i)\n",
    "\n",
    "for i in TestIMG_ls:\n",
    "    i = re.sub(\"IMG\", \"LBL\", i, count = 2)\n",
    "    TestLBL_ls.append(i)\n",
    "\n",
    "#plot a sample\n",
    "test = BuildingDataset(TestIMG_ls, TestLBL_ls, transform = None)\n",
    "test_pair = random.choice(test)\n",
    "\n",
    "assert len(test.png_dir) == len(test.lbl_dir)\n",
    "\n",
    "print(type(test_pair[0]))\n",
    "\n",
    "fig = plt.figure()\n",
    "ax = fig.add_subplot(1, 2, 1)\n",
    "RGB = plt.imshow(test_pair[0].cpu().numpy().transpose(1, 2, 0))\n",
    "ax.set_title('RGB')\n",
    "\n",
    "ax = fig.add_subplot(1, 2, 2)\n",
    "LBL = plt.imshow(test_pair[1].cpu().numpy().transpose(1, 2, 0))\n",
    "ax.set_title('LBL')\n",
    "\n",
    "assert len(test.png_dir) == len(test.lbl_dir)\n",
    "\n",
    "print(\"Concatenated TESTING images: {0} and labels pair: {0} :\".format(len(test.png_dir)))\n",
    "#for x, y in zip(test.png_dir, test.lbl_dir):\n",
    "#    print(f\"Image: {x}\", f\"Label: {y}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "1665b11d-d865-42ee-908d-870a28c73332",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8757105\n"
     ]
    }
   ],
   "source": [
    "qubvelEB1_UNet = smp.Unet(encoder_name = 'efficientnet-b1', encoder_depth = 5, encoder_weights = None, decoder_use_batchnorm = True, decoder_channels = (256, 128, 64, 32, 16), decoder_attention_type = None, in_channels = 3, classes = 1, activation = \"sigmoid\", aux_params = None)\n",
    "qubvel_weights = os.path.abspath(\"/home/chris/Dropbox/HOTOSM/qubvel_UNet/weights/stage3/effb1-f0/checkpoints/best.pth\")\n",
    "qubvel_weights = torch.load(qubvel_weights, map_location = device)\n",
    "qubvelEB1_UNet.load_state_dict(qubvel_weights[\"state_dict\"])\n",
    "qubvelEB1_UNet.to(device = device)\n",
    "\n",
    "print(sum(p.numel() for p in qubvelEB1_UNet.parameters() if p.requires_grad == True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "d1a09bd6-3764-4aab-8692-3164a0bb60a5",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████| 435/435 [01:12<00:00,  5.98it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Per Window statistics: mean Dice Score = 0.5098835825920105, mean Precision = 0.8002525568008423, mean Recall = 0.4331878125667572, mean Specificity = 0.9904452562332153, mean OA = 0.8365511298179626, mean IoU = 0.39099833369255066\n",
      "\n",
      "\n",
      "GLOBAL statistics: Dice Score = 0.4269979884293492, Precision = 0.8950784614765956, Recall = 0.2803757455255547, Specificity = 0.9908803074975654, OA = 0.8365511488640446, IoU = 0.27145419095999085\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "# Loading testing batch\n",
    "test_loader = DataLoader(test, batch_size = 1, shuffle = True)\n",
    "qubvelEB1_UNet.to(device = device)\n",
    "\n",
    "idx = []\n",
    "png_ls = []\n",
    "Dice_ls = []\n",
    "TP_ls = []\n",
    "TN_ls = []\n",
    "FN_ls = []\n",
    "FP_ls = []\n",
    "Precision_ls = []\n",
    "Recall_ls = []\n",
    "Specificity_ls = []\n",
    "OA_ls = []\n",
    "IoU_ls = []\n",
    "\n",
    "for i, (img, lbl) in tqdm(enumerate(test_loader), total = len(test_loader)):\n",
    "    img = img.to(device = device)\n",
    "    lbl = lbl.to(device = device)\n",
    "\n",
    "    with torch.no_grad():\n",
    "        prediction = qubvelEB1_UNet(img)\n",
    "        prediction = prediction.squeeze(0)\n",
    "        lbl = lbl.squeeze(0)\n",
    "\n",
    "        pred_BOOL = (prediction > 0.001).to(torch.float32)\n",
    "\n",
    "        # Booleanise\n",
    "        pred_BOOL = pred_BOOL.bool()\n",
    "        lbl = lbl.bool()\n",
    "\n",
    "        # 1st order statistics\n",
    "        TP = (pred_BOOL * lbl).sum(dim = [0, 1, 2])\n",
    "        TN = ((~pred_BOOL) * (~lbl)).sum(dim = [0, 1, 2])\n",
    "        FN = ((~pred_BOOL) * lbl).sum(dim = [0, 1, 2])\n",
    "        FP = (pred_BOOL * (~lbl)).sum(dim = [0, 1, 2])\n",
    "\n",
    "        # 2nd order statistics\n",
    "        Precision = (TP / (TP + FP))\n",
    "        Recall = (TP / (TP + FN)) #True Positive Rate / Sensitivity\n",
    "        Specificity = (TN / (FP + TN)) #True Negative Rate\n",
    "        OA = ((TP + TN) /(TP + TN + FP + FN))\n",
    "        IoU = (TP / (TP + FN + FP))\n",
    "        Dice = (2 * ((Precision * Recall) / (Precision + Recall)))\n",
    "\n",
    "        TP = TP.detach().cpu().numpy()\n",
    "        TN = TN.detach().cpu().numpy()\n",
    "        FN = FN.detach().cpu().numpy()\n",
    "        FP = FP.detach().cpu().numpy()\n",
    "        Precision = Precision.detach().cpu().numpy()\n",
    "        Recall = Recall.detach().cpu().numpy()\n",
    "        Specificity = Specificity.detach().cpu().numpy()\n",
    "        OA = OA.detach().cpu().numpy()\n",
    "        IoU = IoU.detach().cpu().numpy()\n",
    "        Dice = Dice.detach().cpu().numpy()\n",
    "\n",
    "        idx.append(i)\n",
    "        TP_ls.append(TP)\n",
    "        TN_ls.append(TN)\n",
    "        FN_ls.append(FN)\n",
    "        FP_ls.append(FP)\n",
    "        Precision_ls.append(Precision)\n",
    "        Recall_ls.append(Recall)\n",
    "        Specificity_ls.append(Specificity)\n",
    "        OA_ls.append(OA)\n",
    "        IoU_ls.append(IoU)\n",
    "        Dice_ls.append(Dice)\n",
    "\n",
    "for j in test.png_dir:\n",
    "    png_ls.append(os.path.basename(j))\n",
    "\n",
    "df = {\"id\": idx,\n",
    "      \"png\": png_ls,\n",
    "      \"Dice_score\": Dice_ls,\n",
    "      \"True_Positive\": TP_ls,\n",
    "      \"True_Negative\": TN_ls,\n",
    "      \"False_Negative\": FN_ls,\n",
    "      \"False_Positive\": FP_ls,\n",
    "      \"Precision\": Precision_ls,\n",
    "      \"Recall\": Recall_ls,\n",
    "      \"Specificity\": Specificity_ls,\n",
    "      \"Overall Accuracy\": OA_ls,\n",
    "      \"IoU\": IoU_ls}\n",
    "    \n",
    "TP_SUM = np.sum(df[\"True_Positive\"])\n",
    "TN_SUM = np.sum(df[\"True_Negative\"])\n",
    "FP_SUM = np.sum(df[\"False_Positive\"])\n",
    "FN_SUM = np.sum(df[\"False_Negative\"])\n",
    "\n",
    "# 2nd order statistics\n",
    "GLOB_Precision = (TP_SUM / (TP_SUM + FP_SUM))\n",
    "GLOB_Recall = (TP_SUM / (TP_SUM + FN_SUM)) #True Positive Rate / Sensitivity\n",
    "GLOB_Specificity = (TN_SUM / (FP_SUM + TN_SUM)) #True Negative Rate\n",
    "GLOB_OA = ((TP_SUM + TN_SUM) / (TP_SUM + TN_SUM + FP_SUM + FN_SUM))\n",
    "GLOB_IoU = (TP_SUM / (TP_SUM + FN_SUM + FP_SUM))\n",
    "GLOB_Dice = (2 * ((GLOB_Precision * GLOB_Recall) / (GLOB_Precision + GLOB_Recall)))\n",
    "\n",
    "results_csv = pd.DataFrame(df)\n",
    "\n",
    "print(\"Per Window statistics: mean Dice Score = {0}, mean Precision = {1}, mean Recall = {2}, mean Specificity = {3}, mean OA = {4}, mean IoU = {5}\".format(np.nanmean(Dice_ls), np.nanmean(Precision_ls), np.nanmean(Recall_ls), np.nanmean(Specificity_ls), np.nanmean(OA_ls), np.nanmean(IoU_ls)))\n",
    "print(\"\\n\")\n",
    "print(\"GLOBAL statistics: Dice Score = {0}, Precision = {1}, Recall = {2}, Specificity = {3}, OA = {4}, IoU = {5}\".format(GLOB_Dice, GLOB_Precision, GLOB_Recall, GLOB_Specificity, GLOB_OA, GLOB_IoU))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "595a866f-72ba-4f0e-a02d-c10f33058b9e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 435 entries, 0 to 434\n",
      "Data columns (total 12 columns):\n",
      " #   Column            Non-Null Count  Dtype \n",
      "---  ------            --------------  ----- \n",
      " 0   id                435 non-null    int64 \n",
      " 1   png               435 non-null    object\n",
      " 2   Dice_score        435 non-null    object\n",
      " 3   True_Positive     435 non-null    object\n",
      " 4   True_Negative     435 non-null    object\n",
      " 5   False_Negative    435 non-null    object\n",
      " 6   False_Positive    435 non-null    object\n",
      " 7   Precision         435 non-null    object\n",
      " 8   Recall            435 non-null    object\n",
      " 9   Specificity       435 non-null    object\n",
      " 10  Overall Accuracy  435 non-null    object\n",
      " 11  IoU               435 non-null    object\n",
      "dtypes: int64(1), object(11)\n",
      "memory usage: 40.9+ KB\n",
      "None count            435\n",
      "unique           435\n",
      "top       0.74003804\n",
      "freq               1\n",
      "Name: Dice_score, dtype: object    id                        png   Dice_score True_Positive True_Negative  \\\n",
      "0   0   DZKN15_IMG_6669-3249.png   0.74003804           585         64540   \n",
      "1   1  KBY15_IMG_28215-22059.png          nan             0         64631   \n",
      "2   2    DZK15_IMG_3420-5814.png    0.2420601           282         63488   \n",
      "3   3   KBY15_IMG_8037-16416.png  0.056694284           130         61080   \n",
      "4   4  KBY15_IMG_29241-22914.png    0.6329499          6658         51156   \n",
      "\n",
      "  False_Negative False_Positive   Precision      Recall Specificity  \\\n",
      "0            277            134  0.81363004   0.6786543   0.9979281   \n",
      "1            719            186         0.0         0.0   0.9971304   \n",
      "2           1691             75    0.789916  0.14292955  0.99882007   \n",
      "3           4096            230   0.3611111  0.03076195  0.99624854   \n",
      "4           7052            670   0.9085699  0.48563093   0.9870721   \n",
      "\n",
      "  Overall Accuracy          IoU  \n",
      "0       0.99372864    0.5873494  \n",
      "1        0.9861908          0.0  \n",
      "2         0.973053   0.13769531  \n",
      "3        0.9339905  0.029174147  \n",
      "4       0.88217163   0.46300417  \n"
     ]
    }
   ],
   "source": [
    "print(results_csv.info(), results_csv[\"Dice_score\"].describe(), results_csv.head())\n",
    "\n",
    "log_path = os.path.abspath(\"/home/chris/Dropbox/HOTOSM/HOTOSM_OAM_codeV2/iter_log\")\n",
    "\n",
    "results_csv.to_csv(os.path.join(log_path, \"BASE_15cmALL\", \"OCCtransfer-UNtrained.csv\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "95518000-85cd-425d-96dd-cb293a5bc0c5",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}