notebooks/test_cityscapes_model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import cv2\n",
    "import sys\n",
    "import os\n",
    "import torch\n",
    "from tqdm import tnrange\n",
    "\n",
    "sys.path.insert(0, '..')\n",
    "from adaptis.inference.adaptis_sampling import get_panoptic_segmentation\n",
    "from adaptis.inference.prediction_model import AdaptISPrediction\n",
    "\n",
    "device = torch.device('cuda')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Init dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from adaptis.data.cityscapes import CityscapesDataset\n",
    "\n",
    "dataset_path = '/share/data/vision-greg/cityscapes'\n",
    "dataset = CityscapesDataset(dataset_path, split='val', with_segmentation=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<All keys matched successfully>"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from adaptis.model.cityscapes.models import get_cityscapes_model\n",
    "\n",
    "model = get_cityscapes_model(num_classes=19, norm_layer=torch.nn.BatchNorm2d, backbone='resnet50', with_proposals=True)\n",
    "pmodel = AdaptISPrediction(model, dataset, device)\n",
    "\n",
    "weights_path = '../experiments/cityscapes/003/checkpoints/proposals_last_checkpoint.pth'\n",
    "pmodel.net.load_state_dict(torch.load(weights_path)['model_state'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test on one image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/share/data/vision-greg/cityscapes/leftImg8bit/test/berlin/berlin_000494_000019_leftImg8bit.png\n"
     ]
    }
   ],
   "source": [
    "proposals_sampling_params = {\n",
    "    'thresh1': 0.4,\n",
    "    'thresh2': 0.5,\n",
    "    'ithresh': 0.3,\n",
    "    'fl_prob': 0.10,\n",
    "    'fl_eps': 0.003,\n",
    "    'fl_blur': 2,\n",
    "    'max_iters': 100\n",
    "}\n",
    "\n",
    "image_path = os.path.join(dataset_path, 'leftImg8bit/test/berlin', 'berlin_000494_000019_leftImg8bit.png')\n",
    "print(image_path)\n",
    "\n",
    "image = cv2.imread(image_path)\n",
    "image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n",
    "\n",
    "pred = get_panoptic_segmentation(pmodel, image,\n",
    "                                 sampling_algorithm='proposals',\n",
    "                                 use_flip=True, **proposals_sampling_params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7fb4509a5790>"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pylab\n",
    "pylab.imshow(dataset.convert_to_coco_format(pred)['annotation'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c268a391960344ca92689496dbacd39a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "interactive(children=(BoundedIntText(value=0, description='ix', max=22), Output()), _dom_classes=('widget-inte…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<function __main__.show(ix)>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def show(ix):\n",
    "    import pylab\n",
    "    pylab.figure(figsize=(20,10))\n",
    "    pylab.imshow((pred['instances_mask'] == ix).astype('float32')[...,None] * 0.5  + image.astype('float32')/255/2)\n",
    "import ipywidgets as widgets\n",
    "widgets.interact(show, ix=widgets.BoundedIntText(min=0, max=len(pred['masks']), value=0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Define evaluation function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "from adaptis.coco.panoptic_metric import PQStat, pq_compute, print_pq_stat\n",
    "\n",
    "def test_model(pmodel, dataset,\n",
    "               sampling_algorithm, sampling_params,\n",
    "               use_flip=False, cut_radius=-1):\n",
    "    pq_stat = PQStat()\n",
    "    categories = dataset._generate_coco_categories()\n",
    "    categories = {x['id']: x for x in categories}\n",
    "\n",
    "    for indx in tnrange(len(dataset)):\n",
    "        sample = dataset.get_sample(indx)\n",
    "        pred = get_panoptic_segmentation(pmodel, sample['image'],\n",
    "                                         sampling_algorithm=sampling_algorithm,\n",
    "                                         use_flip=use_flip, cut_radius=cut_radius, **sampling_params)\n",
    "        \n",
    "        \n",
    "        coco_sample = dataset.convert_to_coco_format(sample)\n",
    "        pred = dataset.convert_to_coco_format(pred)\n",
    "\n",
    "        pq_stat = pq_compute(pq_stat, pred, coco_sample, categories)\n",
    "    \n",
    "    print_pq_stat(pq_stat, categories)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test proposals-based point sampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0fe3f9d4aaf241a491a5357744397837",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=500), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "          |     PQ      SQ      RQ     N\n",
      "--------------------------------------\n",
      "All       |  53.14   79.07   65.69    19\n",
      "Things    |  40.37   77.57   51.92     8\n",
      "Stuff     |  62.43   80.16   75.70    11\n"
     ]
    }
   ],
   "source": [
    "test_model(pmodel, dataset,\n",
    "           sampling_algorithm='proposals',\n",
    "           sampling_params=proposals_sampling_params,\n",
    "           use_flip=True)"
   ]
  }
 ],
 "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}