evaluation/DeepFashion eval.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "271cc6a4",
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import os.path as osp\n",
    "import sys\n",
    "from types import SimpleNamespace\n",
    "\n",
    "import numpy as np\n",
    "import torch\n",
    "import torch.nn.functional as F\n",
    "import torchvision.transforms as transforms\n",
    "from PIL import Image\n",
    "from sklearn.metrics import normalized_mutual_info_score\n",
    "from sklearn.metrics import pair_confusion_matrix\n",
    "from torch.utils.data import Dataset\n",
    "import matplotlib.pyplot as plt\n",
    "from tqdm import tqdm\n",
    "\n",
    "from models.model_factory import model_generator\n",
    "\n",
    "\n",
    "def pil_loader(path, type):\n",
    "    with open(path, 'rb') as f:\n",
    "        img = Image.open(f)\n",
    "        return img.convert(type)\n",
    "\n",
    "\n",
    "df_parts = np.array([[0, 0, 0],\n",
    "                     [0, 100, 0],\n",
    "                     [16, 78, 139],\n",
    "                     [50, 205, 50],\n",
    "                     [70, 130, 180],\n",
    "                     [127, 255, 212],\n",
    "                     [144, 238, 144],\n",
    "                     [211, 211, 211],\n",
    "                     [220, 220, 220],\n",
    "                     [245, 222, 179],\n",
    "                     [250, 235, 215],\n",
    "                     [255, 0, 0],\n",
    "                     [255, 140, 0],\n",
    "                     [255, 250, 205],\n",
    "                     [255, 250, 250],\n",
    "                     [255, 255, 0]], dtype=int)\n",
    "\n",
    "\n",
    "def adjusted_rand_score_overflow(labels_true, labels_pred):\n",
    "    (tn, fp), (fn, tp) = pair_confusion_matrix(labels_true, labels_pred)\n",
    "\n",
    "    # Special cases: empty data or full agreement\n",
    "    if fn == 0 and fp == 0:\n",
    "        return 1.0\n",
    "    (tn, fp), (fn, tp) = (tn / 1e8, fp / 1e8), (fn / 1e8, tp / 1e8)\n",
    "    return 2. * (tp * tn - fn * fp) / ((tp + fn) * (fn + tn) +\n",
    "                                       (tp + fp) * (fp + tn))\n",
    "\n",
    "class DF(Dataset):\n",
    "    def __init__(self, split, root='/tmp/datasets', transforms=None):\n",
    "        super().__init__()\n",
    "\n",
    "        self.root = root\n",
    "        self.split = split\n",
    "        self.transforms = transforms\n",
    "\n",
    "        self.image_dir = osp.join(self.root, 'DeepFashion/In-shop Clothes Retrieval Benchmark/')\n",
    "\n",
    "        if self.split == 'train':\n",
    "            self.files = self.only_file_names(json.load(open(f'{self.image_dir}/Anno/segmentation/DeepFashion_segmentation_train.json', 'r'))['images'])\n",
    "        elif self.split == 'test':\n",
    "            val_files = [f'{self.image_dir}/Anno/segmentation/{f}' for f in ['DeepFashion_segmentation_query.json', 'DeepFashion_segmentation_gallery.json']]\n",
    "            self.files = self.only_file_names(json.load(open(val_files[0], 'r'))['images']) + self.only_file_names(json.load(open(val_files[1], 'r'))['images'])\n",
    "        else:\n",
    "            raise ValueError(\"Invalid split name: {}\".format(self.split))\n",
    "        self.num_imgs = len(self.files)\n",
    "        print('%d images' % self.num_imgs)\n",
    "\n",
    "        sum_colors = df_parts.sum(1)\n",
    "        self.col2idx = -np.ones(sum_colors.max() + 1, dtype=np.uint8)\n",
    "        for i in range(sum_colors.shape[0]):\n",
    "            self.col2idx[sum_colors[i]] = i\n",
    "\n",
    "    @staticmethod\n",
    "    def only_file_names(lst):\n",
    "        return [e['file_name'] for e in lst]\n",
    "\n",
    "    def __getitem__(self, index):\n",
    "        path = self.files[index]\n",
    "        image = pil_loader(self.image_dir + '/Anno/segmentation/' + path, 'RGB')\n",
    "        seg = pil_loader(self.image_dir + '/Anno/segmentation/' + path.replace('.jpg', '_segment.png'), 'RGB')\n",
    "        seg = np.array(seg).astype(np.uint8)\n",
    "        seg = self.col2idx[seg.sum(2)]\n",
    "        assert -1 not in seg\n",
    "\n",
    "        mask = np.array(pil_loader(self.image_dir + '/supervisedlabels/' + path + '.png', 'L'))\n",
    "        image = self.transforms[0](image)\n",
    "        seg = self.transforms[1](Image.fromarray(seg))\n",
    "        mask = self.transforms[1](Image.fromarray(mask))\n",
    "        return {'path': path, 'image': image, 'mask': mask, 'seg': seg}\n",
    "\n",
    "    def __len__(self):\n",
    "        return len(self.files)\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "class To_tensor():\n",
    "    def __call__(self, img):\n",
    "        img = np.array(img, np.float32, copy=False)\n",
    "        if len(img.shape) == 2:\n",
    "            img = np.expand_dims(img, 2)\n",
    "        img = torch.from_numpy(img)\n",
    "        img = img.permute((2, 0, 1)).contiguous()\n",
    "        return img\n",
    "\n",
    "from datasets import deepfashion\n",
    "import torch.nn as nn\n",
    "from torch.utils.data import DataLoader\n",
    "from datasets.lit_dataset import LitDataset\n",
    "def get_fg_nmi_ari(path, is_scops, num_parts=4, mask_type='sup_mask'):\n",
    "    opts = SimpleNamespace()\n",
    "    opts.input_size = 256\n",
    "    opts.split = 'test'\n",
    "    opts.unsup_mask = mask_type\n",
    "    opts.unsup_mask2 = False\n",
    "    opts.unsup_train = False\n",
    "    opts.model = \"DeepLab50_2branch\"\n",
    "    opts.dataset_root = '/work/subha/datasets/'\n",
    "    opts.single_class = None\n",
    "    opts.three_classes = None\n",
    "    opts.restore_from = 'None'\n",
    "    opts.num_parts = num_parts\n",
    "    opts.gpu = 0\n",
    "    test_dataset = deepfashion.DFDataset(opts)\n",
    "    test_loader = DataLoader(LitDataset(test_dataset), batch_size=8, shuffle=False, num_workers=5, drop_last=False)\n",
    "    interp = nn.Upsample(size=(opts.input_size, opts.input_size), mode='bilinear', align_corners=True)\n",
    "    model = model_generator(opts, add_bg_mask=is_scops)\n",
    "    model.load_state_dict(torch.load(path) if is_scops else torch.load(path)['model_state_dict'])\n",
    "    model.cuda()\n",
    "    model.eval()\n",
    " \n",
    "    gts = []\n",
    "    preds = []\n",
    "    for batch in tqdm(test_loader):\n",
    "        images_cpu = batch['img']\n",
    "        gt = batch['seg']\n",
    "        images = images_cpu.cuda(opts.gpu)\n",
    "        _, _, pred_low, _ = model(images)\n",
    "        pred = interp(pred_low)\n",
    "\n",
    "        pred_argmax = pred.argmax(dim=1)\n",
    "        gts.append(gt.type(torch.int8))\n",
    "        preds.append(pred_argmax.cpu().type(torch.int8))\n",
    "\n",
    "    gts = torch.cat(gts, 0).flatten()\n",
    "    preds = torch.cat(preds, 0).flatten()\n",
    "\n",
    "    preds = preds[gts != 0]\n",
    "    gts = gts[gts != 0]\n",
    "\n",
    "    fgari = adjusted_rand_score_overflow(preds, gts)*100\n",
    "    fgnmi = normalized_mutual_info_score(preds, gts)*100\n",
    "    return fgnmi, fgari\n",
    "\n",
    "def get_full_nmi_ari(data_loader, model, is_scops, gpu=0):\n",
    "    all_nmi_preds = []\n",
    "    all_nmi_preds_w_bg = []\n",
    "    all_nmi_gts = []\n",
    "    all_nmi_gts_w_bg = []\n",
    "\n",
    "    # iterating the data loader, landmarks shape: [N, num_landmarks, 4], column first\n",
    "    # bbox shape: [N, 5]\n",
    "    for i, batch in enumerate(tqdm(data_loader)):\n",
    "\n",
    "        with torch.no_grad():\n",
    "\n",
    "            seg = batch['seg']\n",
    "            fg_mask = batch['mask'].cuda(gpu)\n",
    "            images = batch['image'].cuda(gpu)\n",
    "            if is_scops:\n",
    "                # preprocessing from scops evaluation code\n",
    "                logit_w_bg = F.interpolate(model(images)[2], size=images.shape[-2:], mode='bilinear', align_corners=False)\n",
    "                logit_w_bg = torch.softmax(logit_w_bg, dim=1)\n",
    "                logit_w_bg /= logit_w_bg.max(dim=3, keepdim=True)[0].max(dim=2, keepdim=True)[0]\n",
    "                logit_w_bg[:,0,:,:] = 0.1\n",
    "                pred = logit_w_bg[:,1:].argmax(1, keepdim=True)\n",
    "                pred_w_bg = logit_w_bg.argmax(1, keepdim=True)\n",
    "            else:\n",
    "                pred = F.interpolate(model(images)[2], size=images.shape[-2:], mode='bilinear', align_corners=False).argmax(1, keepdim=True)\n",
    "                pred_w_bg = fg_mask * (pred + 1)\n",
    "                if i == 0:\n",
    "                    # print(seg.shape, pred.shape, pred_w_bg.shape)\n",
    "                    plt.imshow(seg.cpu().numpy().squeeze()*50)\n",
    "                    plt.show()\n",
    "                    plt.imshow(pred.cpu().numpy().squeeze()*50)\n",
    "                    plt.show()\n",
    "                    plt.imshow(pred_w_bg.cpu().numpy().squeeze()*50)\n",
    "                    plt.show()\n",
    "                    plt.imshow(fg_mask.cpu().numpy().squeeze())\n",
    "                    plt.show()\n",
    "\n",
    "\n",
    "            all_nmi_preds_w_bg.append(pred_w_bg.cpu().view(-1).numpy())\n",
    "            all_nmi_gts_w_bg.append(seg.type(torch.int8).cpu().view(-1).numpy())\n",
    "\n",
    "    all_nmi_preds_w_bg = np.concatenate(all_nmi_preds_w_bg, axis=0)\n",
    "    all_nmi_gts_w_bg = np.concatenate(all_nmi_gts_w_bg, axis=0)\n",
    "\n",
    "    nmi2 = normalized_mutual_info_score(all_nmi_gts_w_bg, all_nmi_preds_w_bg) * 100\n",
    "    ari2 = adjusted_rand_score_overflow(all_nmi_gts_w_bg, all_nmi_preds_w_bg) * 100\n",
    "    return nmi2, ari2\n",
    "\n",
    "\n",
    "def eval_df(path=None, is_scops=False, num_parts=4, req_label=1, image_size=256):\n",
    "    # define data transformation (no crop)\n",
    "    data_transforms = [transforms.Compose([\n",
    "        transforms.Resize(size=(image_size)),\n",
    "        To_tensor(),\n",
    "        transforms.Normalize(mean=(104.00698793, 116.66876762, 122.67891434), std=(1, 1, 1))\n",
    "    ]),\n",
    "        transforms.Compose([\n",
    "            transforms.Resize(size=(image_size)),\n",
    "            To_tensor()\n",
    "        ])]\n",
    "\n",
    "    # define dataset and loader\n",
    "    data = DF(split='test', transforms=data_transforms, root='../data')\n",
    "    loader = torch.utils.data.DataLoader(data, batch_size=1, shuffle=False, num_workers=0, pin_memory=False, drop_last=False)\n",
    "    args2 = SimpleNamespace(dataset_root='/work/subha/datasets/', use_lab=True, single_class=None,\n",
    "                            gpu=0, batch_size=6, input_size=image_size, split='test', model=\"DeepLab50_2branch\",\n",
    "                            num_parts=num_parts, num_classes=15, restore_from='None', unsup_mask=True)\n",
    "\n",
    "    model = model_generator(args2, add_bg_mask=is_scops)\n",
    "    model.load_state_dict(torch.load(path) if is_scops else torch.load(path)['model_state_dict'])\n",
    "    model.cuda()\n",
    "    model.eval()\n",
    "\n",
    "    full_nmi, full_ari = get_full_nmi_ari(loader, model, is_scops=is_scops)\n",
    "    fg_nmi, fg_ari = get_fg_nmi_ari(path, is_scops, num_parts=4, mask_type='sup')\n",
    "    return fg_nmi, fg_ari, full_nmi, full_ari\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "44885f10",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6922 images\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\r",
      "  0%|                                                                                                                                                                                | 0/6922 [00:00<?, ?it/s]"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 6922/6922 [14:26<00:00,  7.99it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6922 images\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 866/866 [02:51<00:00,  5.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "FG-NMI: 44.54 FG-ARI: 46.20 Full-NMI: 68.10 Full-ARI: 90.60\n"
     ]
    }
   ],
   "source": [
    "is_scops = False\n",
    "path = '../checkpoints/DeepFashion/model_100000.pth'\n",
    "fg_nmi, fg_ari, full_nmi, full_ari = eval_df(path=path, is_scops=is_scops)\n",
    "print(f'FG-NMI: {fg_nmi:2.2f} FG-ARI: {fg_ari:2.2f} Full-NMI: {full_nmi:2.2f} Full-ARI: {full_ari:2.2f}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8e0fcd66",
   "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": "3.8.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}