3.Saliency_Maps.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import torch\n",
    "import torchvision\n",
    "import torch.nn as nn\n",
    "import torch.optim as optim\n",
    "\n",
    "from nn_interpretability.interpretation.saliency_map.saliency_map import SaliencyMap\n",
    "from nn_interpretability.model.model_trainer import ModelTrainer\n",
    "from nn_interpretability.model.model_repository import ModelRepository\n",
    "from nn_interpretability.visualization.mnist_visualizer import MnistVisualizer\n",
    "from nn_interpretability.dataset.mnist_data_loader import MnistDataLoader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_name = 'model_am.pt'\n",
    "\n",
    "train = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "mnist_data_loader = MnistDataLoader()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "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": [
    "MnistVisualizer.show_dataset_examples(mnist_data_loader.trainloader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = ModelRepository.get_am_classifier(model_name)\n",
    "\n",
    "if train:\n",
    "    criterion = nn.CrossEntropyLoss()\n",
    "    optimizer = optim.Adam(model.parameters(), lr=0.0005)\n",
    "\n",
    "    model.train()\n",
    "    ModelTrainer.train(model, criterion, optimizer, trainloader)\n",
    "    \n",
    "    ModelRepository.save(model, model_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# I. Saliency Maps"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Saliency Maps for each class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "images = []\n",
    "img_shape = (28, 28)\n",
    "\n",
    "for i in range(10):\n",
    "    img = mnist_data_loader.get_image_for_class(i)\n",
    "    \n",
    "    interpretor = SaliencyMap(model, [], img_shape, None)\n",
    "    endpoint = interpretor.interpret(img)\n",
    "    endpoint = MnistVisualizer.normalize(endpoint)\n",
    "\n",
    "    images.append(endpoint.reshape(img_shape))\n",
    "\n",
    "MnistVisualizer.display_images_for_each_class(images)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Saliency map for each class multiplied by the input"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "images = []\n",
    "img_shape = (28, 28)\n",
    "\n",
    "for i in range(10):\n",
    "    img = mnist_data_loader.get_image_for_class(i)\n",
    "    \n",
    "    interpretor = SaliencyMap(model, [], img_shape, None, True)\n",
    "    endpoint = interpretor.interpret(img)\n",
    "    endpoint = MnistVisualizer.normalize(endpoint)\n",
    "   \n",
    "    images.append(endpoint.detach().numpy().reshape(img_shape))\n",
    "\n",
    "MnistVisualizer.display_images_for_each_class(images)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}