CNN and CNN Visualization.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "PkvLxjy_lX5K"
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "import torchvision\n",
    "import torchvision.transforms as transforms\n",
    "import torchvision.datasets as dsets\n",
    "import random as rand\n",
    "import numpy\n",
    "import numpy as np\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt \n",
    "from torch.autograd import Variable\n",
    "from sklearn.metrics import confusion_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 52
    },
    "colab_type": "code",
    "id": "87S8r8a1lX5W",
    "outputId": "5959c84a-ce16-49d8-b4bb-811f82671dbe"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Files already downloaded and verified\n",
      "Files already downloaded and verified\n"
     ]
    }
   ],
   "source": [
    "transform = transforms.Compose(\n",
    "    [transforms.ToTensor()])\n",
    "\n",
    "trainset = torchvision.datasets.CIFAR10(root='./data', train=True,\n",
    "                                        download=True, transform=transform)\n",
    "\n",
    "\n",
    "testset = torchvision.datasets.CIFAR10(root='./data', train=False,\n",
    "                                       download=True, transform=transform)\n",
    " \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "ffeiQ4FTlX5i"
   },
   "outputs": [],
   "source": [
    "'''\n",
    "STEP 2: MAKING DATASET ITERABLE\n",
    "'''\n",
    "\n",
    "batch_size = 100\n",
    "n_iters = 15000\n",
    "num_epochs = n_iters / (len(trainset) / batch_size)\n",
    "num_epochs = int(num_epochs)\n",
    "\n",
    "train_loader = torch.utils.data.DataLoader(dataset=trainset, \n",
    "                                           batch_size=batch_size, \n",
    "                                           shuffle=True)\n",
    "\n",
    "test_loader = torch.utils.data.DataLoader(dataset=testset, \n",
    "                                          batch_size=batch_size, \n",
    "                                          shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "tOzhGVaolq9b",
    "outputId": "12965ecb-9c43-4707-c5c0-33ed4bc3fe99"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "30"
      ]
     },
     "execution_count": 34,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "num_epochs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "R2KDanxslX5r"
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "\n",
    "class CNN(nn.Module):\n",
    "    \"\"\"CNN.\"\"\"\n",
    "    def __init__(self):\n",
    "        super(CNN, self).__init__()\n",
    "\n",
    "        \n",
    "\n",
    "        # Conv Layer block 1\n",
    "        self.conv1 = nn.Conv2d(in_channels=3, out_channels=32, kernel_size=3, padding=1)\n",
    "        self.norm1 = nn.BatchNorm2d(32)\n",
    "        self.relu = nn.ReLU(inplace=True)\n",
    "        self.conv2 = nn.Conv2d(in_channels=32, out_channels=64, kernel_size=3, padding=1)\n",
    "        #nn.ReLU(inplace=True)\n",
    "        self.pool = nn.MaxPool2d(kernel_size=2, stride=2)\n",
    "\n",
    "        # Conv Layer block 2\n",
    "        self.conv3 = nn.Conv2d(in_channels=64, out_channels=128, kernel_size=3, padding=1)\n",
    "        self.norm2 = nn.BatchNorm2d(128)\n",
    "        #nn.ReLU(inplace=True)\n",
    "        self.conv4 = nn.Conv2d(in_channels=128, out_channels=128, kernel_size=3, padding=1)\n",
    "        #nn.ReLU(inplace=True)\n",
    "        #nn.MaxPool2d(kernel_size=2, stride=2)\n",
    "        #nn.Dropout2d(p=0.05)\n",
    "\n",
    "        # Conv Layer block 3\n",
    "        self.conv5 = nn.Conv2d(in_channels=128, out_channels=256, kernel_size=3, padding=1)\n",
    "        self.norm3 =nn.BatchNorm2d(256)\n",
    "        #nn.ReLU(inplace=True)\n",
    "        self.conv6 = nn.Conv2d(in_channels=256, out_channels=256, kernel_size=3, padding=1)\n",
    "        #nn.ReLU(inplace=True)\n",
    "        #nn.MaxPool2d(kernel_size=2, stride=2)\n",
    "        \n",
    "\n",
    "\n",
    "        \n",
    "        self.drop = nn.Dropout(p=0.5)\n",
    "        self.fc1 = nn.Linear(4096, 1024)\n",
    "        #nn.ReLU(inplace=True)\n",
    "        self.fc2 = nn.Linear(1024, 512)\n",
    "        #nn.ReLU(inplace=True)\n",
    "        #nn.Dropout(p=0.5)\n",
    "        self.fc3 = nn.Linear(512, 10)\n",
    "        \n",
    "\n",
    "\n",
    "    def forward(self, x):\n",
    "        \"\"\"Perform forward.\"\"\"\n",
    "        \n",
    "        # conv layer1 \n",
    "        out = self.conv1(x)\n",
    "        out = self.norm1(out)\n",
    "        out = self.relu(out)\n",
    "        out = self.conv2(out)\n",
    "        out = self.relu(out)\n",
    "        out = self.pool(out)\n",
    "\n",
    "        #convolution block2\n",
    "        out = self.conv3(out)\n",
    "        out = self.norm2(out)\n",
    "        out = self.relu(out)\n",
    "        out = self.conv4(out)\n",
    "        out = self.relu(out)\n",
    "        out = self.pool(out)\n",
    "        \n",
    "        #convolution block3\n",
    "        out = self.conv5(out)\n",
    "        out = self.norm3(out)\n",
    "        out = self.relu(out)\n",
    "        out = self.conv6(out)\n",
    "        out = self.relu(out)\n",
    "        out = self.pool(out)\n",
    "\n",
    "        # flatten\n",
    "        out = out.view(out.size(0), -1)\n",
    "        \n",
    "        #Linear Function\n",
    "        out = self.drop(out)\n",
    "        out = self.fc1(out)\n",
    "        out = self.relu(out)\n",
    "        out = self.fc2(out)\n",
    "        out = self.relu(out)\n",
    "        out = self.drop(out)\n",
    "        out = self.fc3(out)\n",
    "        \n",
    "    \n",
    "\n",
    "        return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "c6RIBTMcowZ6"
   },
   "outputs": [],
   "source": [
    "model = CNN()\n",
    "'''\n",
    "STEP 4: INSTANTIATE MODEL CLASS\n",
    "'''\n",
    "\n",
    "#######################\n",
    "#  USE GPU FOR MODEL  #\n",
    "#######################\n",
    "\n",
    "if torch.cuda.is_available():\n",
    "    model.cuda()\n",
    "\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=0.0001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 562
    },
    "colab_type": "code",
    "id": "DqfWEE-Jo8p7",
    "outputId": "1b1ed4ef-5ed2-481f-eb14-72ae4810327c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration: 1 | Loss: 1.4609868088960647 \n",
      "Iteration: 2 | Loss: 0.988293151974678 \n",
      "Iteration: 3 | Loss: 0.7905496588945389 \n",
      "Iteration: 4 | Loss: 0.6633407422304154 \n",
      "Iteration: 5 | Loss: 0.5667543399930001 \n",
      "Iteration: 6 | Loss: 0.4919413368701935 \n",
      "Iteration: 7 | Loss: 0.42495089140534403 \n",
      "Iteration: 8 | Loss: 0.3661731256842613 \n",
      "Iteration: 9 | Loss: 0.3137505283951759 \n",
      "Iteration: 10 | Loss: 0.273065834030509 \n",
      "Iteration: 11 | Loss: 0.23411874106526376 \n",
      "Iteration: 12 | Loss: 0.20054009352624416 \n",
      "Iteration: 13 | Loss: 0.17794297055900096 \n",
      "Iteration: 14 | Loss: 0.16084255891293286 \n",
      "Iteration: 15 | Loss: 0.13731075440347196 \n",
      "Iteration: 16 | Loss: 0.12781605553999542 \n",
      "Iteration: 17 | Loss: 0.11196734678000211 \n",
      "Iteration: 18 | Loss: 0.08650543441064656 \n",
      "Iteration: 19 | Loss: 0.07463961458392442 \n",
      "Iteration: 20 | Loss: 0.0671950440686196 \n",
      "Iteration: 21 | Loss: 0.06661647062562406 \n",
      "Iteration: 22 | Loss: 0.06622552843205631 \n",
      "Iteration: 23 | Loss: 0.058581727757118646 \n",
      "Iteration: 24 | Loss: 0.05880536261666566 \n",
      "Iteration: 25 | Loss: 0.05955860868934542 \n",
      "Iteration: 26 | Loss: 0.05290618768706918 \n",
      "Iteration: 27 | Loss: 0.05124657279904932 \n",
      "Iteration: 28 | Loss: 0.049650705613661555 \n",
      "Iteration: 29 | Loss: 0.0493256776612252 \n",
      "Iteration: 30 | Loss: 0.049322208805009725 \n",
      "==> Finished Training ...\n"
     ]
    }
   ],
   "source": [
    "for epoch in range(num_epochs):\n",
    "    loss_values = []\n",
    "    running_loss = 0.0\n",
    "    for i, (images,labels) in enumerate(train_loader):\n",
    "        # get the inputs\n",
    "        #inputs, labels = data\n",
    "\n",
    "        if torch.cuda.is_available():\n",
    "            images = Variable(images.cuda())\n",
    "            labels = Variable(labels.cuda())\n",
    "        else:\n",
    "            images = Variable(images)\n",
    "            labels = Variable(labels)\n",
    "\n",
    "        # wrap them in Variable\n",
    "        #inputs, labels = Variable(inputs), Variable(labels)\n",
    "\n",
    "        # zero the parameter gradients\n",
    "        optimizer.zero_grad()\n",
    "\n",
    "        # forward + backward + optimize\n",
    "        outputs = model(images)\n",
    "        loss = criterion(outputs, labels)\n",
    "        loss.backward()\n",
    "\n",
    "        if epoch > 16:\n",
    "            for group in optimizer.param_groups:\n",
    "                for p in group['params']:\n",
    "                    state = optimizer.state[p]\n",
    "                    if state['step'] >= 1024:\n",
    "                        state['step'] = 1000\n",
    "        optimizer.step()\n",
    "\n",
    "\n",
    "        # print statistics\n",
    "        running_loss += loss.item()\n",
    "\n",
    "    # Normalizing the loss by the total number of train batches\n",
    "    running_loss /= len(train_loader)\n",
    "    loss_values.append(running_loss)\n",
    "    \n",
    "\n",
    "    print(\"Iteration: {0} | Loss: {1} \".format(epoch+1, running_loss))\n",
    "\n",
    "    \n",
    "print('==> Finished Training ...')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "qh3ugaWpuFdw",
    "outputId": "72e3cf47-b8d0-4368-870b-9370449de43c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on total test data: 80%\n"
     ]
    }
   ],
   "source": [
    "correct = 0\n",
    "total = 0\n",
    "# Iterate through test dataset\n",
    "for images, labels in test_loader:\n",
    "    \n",
    "    if torch.cuda.is_available():\n",
    "        images = Variable(images.cuda())\n",
    "    else:\n",
    "        images = Variable(images)\n",
    "                \n",
    "    # Forward pass only to get logits/output\n",
    "    outputs = model(images)\n",
    "                \n",
    "    # Get predictions from the maximum value\n",
    "    _, predicted = torch.max(outputs.data, 1)\n",
    "                \n",
    "    # Total number of labels\n",
    "    total += labels.size(0)\n",
    "                \n",
    "              \n",
    "    # Total correct predictions\n",
    "    if torch.cuda.is_available():\n",
    "        correct += (predicted.cpu() == labels.cpu()).sum()\n",
    "    else:\n",
    "        correct += (predicted == labels).sum()\n",
    "            \n",
    "accuracy = 100 * correct / total\n",
    "print(\"Accuracy on total test data: {}%\".format(accuracy.item()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 284
    },
    "colab_type": "code",
    "id": "Pz7Kyv6fJI__",
    "outputId": "727b2e10-36da-4c78-e8fe-6abdd99196c5"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f4aa6961978>"
      ]
     },
     "execution_count": 58,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPsAAAD5CAYAAADhukOtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjIsIGh0\ndHA6Ly9tYXRwbG90bGliLm9yZy8li6FKAAAdHUlEQVR4nO2de2yc13nmn3eGwzspkrqQulq+KL4k\nkeUso0iJHSROk3jdFI6LxohbBG5rVEZbFxsg+4c3C2yywAKbBJsEWew2XTlx47hx7DRXd+Gt4zoJ\nvN42jinHF8lyLMm6WBRFiiLF+2U48+4fM0Jl4zyH9JAcyjnPDxA0PO+c73u/M9/7ffOdZ973mLtD\nCPHbT2alHRBCVAcFuxCJoGAXIhEU7EIkgoJdiERQsAuRCDWL6WxmNwH4GoAsgG+4+xdi7+/oyPjm\nzWSXMQnQrGIfl5KLwwuxFFRdcK5A4q7Ex5MnCxgaKgZP1YqD3cyyAP4ngA8DOAngGTN7xN1fYn02\nb67BY4+uCdoKkcHIXiTBnuLXoGzkEleofsgsGcUq7y92ftM+FeznYzcPUttizt+dAA67+6vuPgvg\nIQC3LGJ7QohlZDHBvhHAaxf8fbLcJoS4CFn2b6ZmtsfMesys5+zZan95EkKcZzHB3gtg8wV/byq3\nvQ533+vu3e7evXp1ik+9QlwcLCb6ngGwzcwuNbNaAJ8E8MjSuCWEWGoqno139zkzuxvAYyhJb/e5\n+4FYH0Pk6rLEM+7L8R0iNjP9ViZT4dhnIuNRvOizKSvzr9IH0ZiixGbqs5HtVTJTvyid3d0fBfDo\nYrYhhKgOeogWIhEU7EIkgoJdiERQsAuRCAp2IRJhUbPxbxZHZdIFuyL9tkphi6FSGW2pWWo/llrK\nq/zc4X4stSwXTQ4j7bGj0p1diERQsAuRCAp2IRJBwS5EIijYhUiEqs7GF90xUQzPMLZl+HUnZ2Hb\nxVIWKe+VzcPmI7Otuchsdr3FUiTCxMaq0pnuSmbcY/uKbS9mq6R0VqXHHJ/FX+LzsRJFIzaGi3BF\nCPEWQsEuRCIo2IVIBAW7EImgYBciERTsQiRCVaW3V3u78If3fDpoG7mCX3ce/NOvBtvfluN98hVV\n6YpLK82Z+mD7t0bX0T7fOHY9tW1oHqG2XW1Hqe2v2g9R23hxhtoYy5E8Q8cxsqtKV5+pRIKt9Jhj\nkt3SJ2YtbdKN7uxCJIKCXYhEULALkQgKdiESQcEuRCIo2IVIhEVJb2Z2DMAYSqvRzLl7d+z9mTlH\n3bmwJNZxkMsMf/7yHwbbf3/zc7TPquwUtW3NnaG2+kye2mqJnDddzNE+kz/upLbjk9z29s/0UVsu\nkvXGMgTrjPtYrLh62psnU2mGXcS21P5fLNmUS81S6OwfdPfBJdiOEGIZ0dd4IRJhscHuAH5qZvvM\nbM9SOCSEWB4W+zX+enfvNbN1AB43s5fd/ckL31C+COwBgLr6tkXuTghRKYu6s7t7b/n/AQA/ArAz\n8J697t7t7t252qbF7E4IsQgqDnYzazKzlvOvAXwEwP6lckwIsbQs5mt8J4AfWSl7qAbAg+7+j7EO\nmXwRDX0TQdvIVa20n//d2mD7tzd8lPaZXsPlk8Yrz1HbJy77NbWNFcJZbz8/tY32aX1tjtoajo9R\n2/1P3UBtHR8IjyHAfSxGMrJyxjMEr204QW2NGZ5hx2TKjFUmk7VlZrkftrRSWW2FRRuZ7AlUlhGX\nqSSbL2KrONjd/VUA11baXwhRXSS9CZEICnYhEkHBLkQiKNiFSAQFuxCJUNWCkygWYePhbLS2F7hE\nNbB7dbC9/iyXJppOcdv4cDu1/e3YbmqrOR6WtXLjXFY596lxauv4B/6Lwqv+x1lq+0qGS46XbTsd\nbC8U+XW9b5jLnptWX01tN3UdoLZ8MXxq5TL8c85GpKaOGj6OMZg8WHQ+HjF5MJbhGMuYrDdum/bw\nNmP7aiJS5GhxmPbRnV2IRFCwC5EICnYhEkHBLkQiKNiFSITqzsY7YAUy0zk8SrvVj4RnzyfX8mtV\nUz9P7mg9yuvTzRyoo7bG4+GZzlM3dtA+2zeeorYzI5dRG3r7qenKvbXUduw/rAq2//6252mf/XUb\nqG1LI5/dfWLgKmp7W+tAsD02C354bA21Ned40s26ej5Tf0l9WNXozPGlt4byzdQWUwWG5ni/mC02\ni884SWbqp4tHaB/d2YVIBAW7EImgYBciERTsQiSCgl2IRFCwC5EI1ZXe4MBcWBLziUnaKzceluta\np3jCQm6UJ1yc3sWr3K46yiW7zHBYdinWcultPM+lPCvyxA9bxZNTsn18AZ6W/xOW8x46wxN8MlM8\nkedAdiu1FRv4+B9qXkdtDMvw8Who4DXo9vVxCTPTGpa1VrdzCW10IpzwBAB1tfy8ykRq4U3N8KQW\nts3Y9kZGGoPtg1N8STTd2YVIBAW7EImgYBciERTsQiSCgl2IRFCwC5EI80pvZnYfgI8BGHD3d5Tb\nOgA8DGArgGMAbnN3nh51HgdQDMs17lxmyDeHr0nZGd4nd5pnNbW/wmWQQj2XoSbeuZ7aGK8c2ERt\nl09wOQlzXOKJse4X4Sy75j4uhR27lW/vsst59t2p4XCGHQC0NYel1NUNXGI9PMCz3sZP86yxXMc0\ntTFqs1xibW/hPp59no9jcQvPpqyv55ltM7PhMCwc5cecZadpvrKlq87zLQA3vaHtHgBPuPs2AE+U\n/xZCXMTMG+zl9daH3tB8C4D7y6/vB/DxJfZLCLHEVPrM3unufeXXp1Fa0VUIcRGz6Ak6Lz1s04dn\nM9tjZj1m1jNb5M80QojlpdJg7zez9QBQ/j9cgwiAu+919253767NNFS4OyHEYqk02B8BcEf59R0A\nfrI07gghlouFSG/fBfABAGvM7CSAzwH4AoDvmdmdAI4DuG0hO/PaGsxtDssrmQNjtF++MXxNosUr\nARjJrgOATJ73K9Zmqa3hRNjHuZ08663pON9e7TGevYYs7+f5SIHCmnC/hsN8X1t/xCWvU3dyeW1D\nO5c3P7FhX7B9R/0J2mdgYwu1fenIGwWhf6X3tfDyYADw9m0ng+1XtnBJ8cAIl1hPd3K59I/f/itq\n66iZoLYfntoRbD9BJDkAaGgM+5Gp4+f9vMHu7rcT04fm6yuEuHjQL+iESAQFuxCJoGAXIhEU7EIk\ngoJdiESoasFJzxjmmsIZZ4Ub+Lph45vCmTytR3mGT4x8c0TWilz+CqvChQjX7eMZauMb+L4mrumi\ntsYjb0xH+FdsNuI/yx7M8ANrOH6O2up/sZba+j7Isw7vz+8Ktn/x6O/SPtk2LmvVHAoXWASAq7/D\nZbTpjWEZ7R93X0r7TG7hn6cV+Dn3wP73UBsixSPr6sJSai5S3LJQCH+e7ovLehNC/BagYBciERTs\nQiSCgl2IRFCwC5EICnYhEqGq0ltmJo/6V8Iyibfy9dcaNoSzyvLNXGYYfg/PXJpt4f1qpiLrdXWG\n120b3M6lsI5dp6lt8Bdcetvcy4tiIlKck0ps2ch13fh4dD3F64jOHOKf2URXOINt6wCXk85dzuW1\n3HjkmCPkhsIFU7Y+QEsw0KKoADD4oUuo7czOWmrzbOS8mgufVzUT/DPLjYQ/M5vkfXRnFyIRFOxC\nJIKCXYhEULALkQgKdiESoaqz8Sg6fHomaPIhPutbf3VbsL12lNfbaniFz7aOv5PPgmfyfNa0ZiI8\nk2wFXjW373Q7tTVE8ngmt7RSW9MrZ3lHUoPO6XpBiCbJeGSmvnaIL7uUGw0ntViBj29HpDagFXm/\nYmukajHZX7GdL61kx8JLaAHx82PL23hCzoYmXq9vuhAOw6FprnbkyPJVvd/nyUS6swuRCAp2IRJB\nwS5EIijYhUgEBbsQiaBgFyIRFrL8030APgZgwN3fUW77PIA/A3Cm/LbPuvujy+XkdFv4mhST3nyc\nL7cTrTNXF5GazoX3t64nshzTPr4zK/J+x26L1Cw7uY7aLns4XLvO63hiTaRsWax0WlzOIxTqeNJQ\nTJazSPKPZ7gfbJvTG/hSU/XFDdTW3sMl3cPv5MlXxzv5Elsg7lsmcg40hSW22Tke0gu5s38LQGih\nra+6+47yv2ULdCHE0jBvsLv7kwB4qVMhxFuCxTyz321mL5jZfWbGfyYmhLgoqDTYvw7gcgA7APQB\n+DJ7o5ntMbMeM+uZLYYLCQghlp+Kgt3d+9294O5FAPcC2Bl5715373b37tpM5DfMQohlpaJgN7ML\npx1vBbB/adwRQiwXC5HevgvgAwDWmNlJAJ8D8AEz2wHAARwDcNeC9uYO5Ek2VEN4aSWAy2FRqSbH\npaZ8E7/G1Q1H5Dwi8Qxfxfc1tY77mF/D67HlGrgsty4i53kuLG1FJa+YvBardxfJiCvmyPJEEbmu\nUBepnzbLM+Kyk9REJbuZNj4eQ1eGax4CwMYHD1Hb6v1cEu1v5fvLkrpxWZ5UiNrhcN06i9StmzfY\n3f32QPM35+snhLi40C/ohEgEBbsQiaBgFyIRFOxCJIKCXYhEqG7BSQNQE96lz/BCeUWibA1dHZYf\nAKCpjS/TM7kuVmCRmlAg8uAcVw0xt54fV1s7z8y7YeMRavtleze1NR0JS1Sxgo0eWxoqklEWyzYD\n2V1MXovJYfWR7AzLR+TSmvD+ptZwP8auCxdFBQA8zn8Znp3hY7xpG8+WOzMaLn45PcrP7xlyDhfr\nuA+6swuRCAp2IRJBwS5EIijYhUgEBbsQiaBgFyIRqiq9FVrrce53tgVtMclrem1YTmjs451ajoxR\n22zzKmqrH+IyTm48nKW2ZpZnvc0d5kOcm+BFD3+2/d3U1hjLRCNSU4xibWVSZBQiy822cHktVgg0\nN8azAGPr0bHMvEIt74IR/nnOruNrxK3a10dtfT/eSG3eFfaxpoF/znMt/Dxl6M4uRCIo2IVIBAW7\nEImgYBciERTsQiRCVWfjs+N5tD8dnrGc3cLrfmXmwlOnrYf4jHtmhCeZZGdbqS1Gvjk8XEPX8GEc\n38rrzNUN8n4eWfpnejWffZ7a2BRsrx3mCTmx5JTsDK/9Fk1qWRWedR+8lvs+18l9bPq7SA23cZ64\nUjRSCy9y5nstP+aJ9Xwav+5l7kfrMX4e5JvDs/9WiCxFNhw+gMws76M7uxCJoGAXIhEU7EIkgoJd\niERQsAuRCAp2IRJhIcs/bQbwbQCdKFUW2+vuXzOzDgAPA9iK0hJQt7n7cHRjc3Mo9p8JmmpneaLD\nme1bg+0Ng7z4W10/L1o2syoiXa3lSRCN/USS4UoN6ga5ZDSzPrL8UzOXoaZyjdRW82w4QeLU+8OS\nHAB0/s5Jajvev5raCmP82JrWhWXRtlr+OTfV8mMefMcGatv401FqQ134FC9Gznxr5J/L2e28Llzd\n6BZqy0QkzDnycc50RE6stvBYFesXV4NuDsBn3P0aALsA/KWZXQPgHgBPuPs2AE+U/xZCXKTMG+zu\n3ufuz5ZfjwE4CGAjgFsA3F9+2/0APr5cTgohFs+bemY3s60ArgPwNIBOdz//c7jTKH3NF0JcpCw4\n2M2sGcAPAHza3V/3kOSldX2DDwtmtsfMesysZ9Yja9AKIZaVBQW7meVQCvTvuPsPy839Zra+bF8P\nIFgF3933unu3u3fXWmQ1BSHEsjJvsJuZobQe+0F3/8oFpkcA3FF+fQeAnyy9e0KIpWIhWW/vA/Ap\nAC+a2XPlts8C+AKA75nZnQCOA7htvg25OzxPZI0aLuPkSdmvuUbeZ/Qjl1JboZ5Lb12/HKe26bXh\nbyZ15/j2cnxzaD7Oh3+uiUuANVNcXqkZCWdeNZ/k2Vqv9fD6aLUTlR1b/mS4zt8UV7UwxU0oRhIV\n59oiUuRIeKutR7kUmZ3h30Brz0XqwtXze2dujvdrGAjbaib49gpEdrYZ/nnNG+zu/hRKq7SF+NB8\n/YUQFwf6BZ0QiaBgFyIRFOxCJIKCXYhEULALkQhVLTgJd3iBLFuTiVx3iBbQ917uftcv+fI4Mclu\npp1nNYV/IwgMbefZSc1beEZWLst9fG/XCWp7/izPAJscWBNsH9xBu+CG6/dT26Fza6mtt5cXCb3y\n0nBh0eYcL8rYN8H1tdgqVEOn+C+1O396NtjuWe77dDcvVpo/yCW7s+/m8lr9KS59ZogcmY384HS6\nM3zuOFdsdWcXIhUU7EIkgoJdiERQsAuRCAp2IRJBwS5EIlRXegOAIpEMGnmm0eQl4SKFGx/n16qm\nV0eobehKXkRxai0fktxEWGJb28P9yL/URm2zzVxQ+lkr9zE7zft1ToULETb08+M6cLaL2sYmIzUI\nZvlx946Es97qczztra2B57011PBClS++m0uYa/eF5bzWV/m++ie5fpXfyP1Yu+kctc10RdYDHAuP\n8eoOnlbYSmTbwTrun+7sQiSCgl2IRFCwC5EICnYhEkHBLkQiVHU2vrCmCUO37A7aJrv4DHPHvnCC\nQdMJPltZbOSJB4XIBHNzL1+CqP6l3vC+OnlSxct3kQJ6AJq7Iv5PR2aEC/waPXUsnMiT4YeFoRGe\n3FGYjpwiOZ4AND4Urgs3Ocq3xxfsAmyOnx+Nw5HlvNaHj61mis/gNx/k585sO092GT/Jk4ayPP8H\ndWR3o7kG2qeYC/sxF1ESdGcXIhEU7EIkgoJdiERQsAuRCAp2IRJBwS5EIswrvZnZZgDfRmlJZgew\n192/ZmafB/BnAM6U3/pZd380tq3MjKPlRFgDys5wyWBwB5NWuKyVjySZ1PKycCjURpbc2RSWVjLn\neM2ypuM8EWayhde7233ZUWp7zypu++8nbw62F9bzgmZ3b3+S2qaL/HNpiRRJy1k44SXv/JSbLEbq\ntBmX+e7d/z5qO9VKpLcxflwbb3yN2t7eFq6tBwBFj0iABX5sfVPhZJ33drxK++Q9XEfx3gfGaJ+F\n6OxzAD7j7s+aWQuAfWb2eNn2VXf/bwvYhhBihVnIWm99APrKr8fM7CAAvhKgEOKi5E09s5vZVgDX\nAXi63HS3mb1gZveZWfsS+yaEWEIWHOxm1gzgBwA+7e6jAL4O4HIAO1C683+Z9NtjZj1m1pPP82db\nIcTysqBgN7McSoH+HXf/IQC4e7+7F9y9COBeADtDfd19r7t3u3t3Lsd/gy2EWF7mDXYzMwDfBHDQ\n3b9yQfv6C952KwC+rIgQYsUxd57FAwBmdj2A/wvgRQDn9Y/PArgdpa/wDuAYgLvKk3mUVbWd/t6u\n24O2uU285tqJj4QlthmyBA4AtL7Ml3ga2c7rdMH4eHRuCNcYm/j5Otpn/VP80SU7wVOh5lp5at5A\ndzijDAAaB8ISVUQVwuhWfs2fWcMlr2IDt7UcCs/9Tq/l41uo47bMLD+AWE0+rwlvs36Q9xm/JLKc\n11Ze23AqkqlY7OWfWc3m8DlyxbpB2ufKlv5g+4N/9Dj6XxoKHtxCZuOfQnipraimLoS4uNAv6IRI\nBAW7EImgYBciERTsQiSCgl2IRKhqwcm5lloMvX9z0DZ6Cb/utBwPyyeN/bxP4wCX5SbfxyWvYoFL\ndoPDLeE+W/i++nfxHxI1nOEFBfNNXBr6xJ/+jNoefPjGYHvzSS5r/cntj1Hb/3rsw9TWcJKfPlt+\nL5yZd/jJrbTPume4j4PX8vHo+hWXUm0uvE0rRGS+PM9GHMmFM9QAwOv4eWARWXFmKHwevNS/hfb5\nzeTWYPu50X+mfXRnFyIRFOxCJIKCXYhEULALkQgKdiESQcEuRCJUVXrLFBy1Y+GMormI1HRmd1jS\nWN3DZbKxTdyWP8czynJtvIhi4XQ4c6lumF8zp1dzyWXtr/kCbJNdvEDhqRlexHLV0fD41o7yTK7B\nfFhSBIBCe7hwJAC0/5KP8RW3ngm2F2/gn/Po82FZFgDW7AxneQHAia18PNY9Fh7HSHIj7vqLn1Db\n7a2HqS3vfIyzwVyyEtORfm+Wmx/gmXK6swuRCAp2IRJBwS5EIijYhUgEBbsQiaBgFyIRqiq92Zyj\n7mw448yKvCBf/emwm2OXcv2k7iz3Y83TXDIavYJnqa3eEZY1RvetoX3W/zOXrvp2cwmwbpgf2z8d\neRu1XdJLxjfP5Z1DY+E17AAABS4ZZfih4ZH/92+C7Xd9kGfs/X3bJdRW/w3uY+uGSKbideHj3vw4\nd/7Hp3dQ28eaf0NtY0V+76yNrFU36+F++ci9uN7CcnQhIvHpzi5EIijYhUgEBbsQiaBgFyIRFOxC\nJMK8s/FmVg/gSQB15fd/390/Z2aXAngIwGoA+wB8yt15ZgeAQl0Go5eG623N8XJsKK0wFWjt4kkr\n2Sm+wdbjfGZ0XQ/3Ymg8PCNsfNUfnPhdfj1dtWmI2uof5Mkd43x3OP7R8Ax/5zP8mI88ezm1ZXlZ\nNfTeyj/uphfDfvztwV20T6YrskZVho9jPrw6GABg0/bTwfbB/g20T3GYj/1TUzxZ5z31r1Hbi7N8\nibB6C9fQy4OrDAUygz/t4SXKgIXd2WcA3Oju16K0tttNZrYLwBcBfNXdrwAwDODOBWxLCLFCzBvs\nXuL8zSRX/ucAbgTw/XL7/QA+viweCiGWhIWuz541s+cADAB4HMARAOfc/fwvE04C2Lg8LgohloIF\nBbu7F9x9B4BNAHYCuGqhOzCzPWbWY2Y9czN8+WIhxPLypmbj3f0cgJ8D2A2gzczOT/BtAtBL+ux1\n9253766p4z9FFUIsL/MGu5mtNbO28usGAB8GcBCloP+D8tvuAMBr+QghVpyFJMKsB3C/mWVRujh8\nz93/t5m9BOAhM/svAH4N4JvzbSg7XUDbb8LCUdNpvuTOyQ+G64hlX+byWpGXcMPQ1VzSaDnBE1Da\nXwnrUDOr+DXTtvNHl46mSWobaW6ntj+48jlq++Pd/xJsv7nuM7RPtov78ZNdf0NtXzr9UWp7/l/e\nEWyffYXrZHU7uRQ561yWa/oxl8pOHOoMtte0R5Z/OsCXePqvT91ObTEJsBiJtLZXwrJo69Ep2scz\n4fEYOPbXtM+8we7uLwC4LtD+KkrP70KItwD6BZ0QiaBgFyIRFOxCJIKCXYhEULALkQjmHlkHZ6l3\nZnYGwPHyn2sA8LVqqof8eD3y4/W81fy4xN2D6ZlVDfbX7disx927V2Tn8kN+JOiHvsYLkQgKdiES\nYSWDfe8K7vtC5MfrkR+v57fGjxV7ZhdCVBd9jRciEVYk2M3sJjP7jZkdNrN7VsKHsh/HzOxFM3vO\nzCKlJpd8v/eZ2YCZ7b+grcPMHjezQ+X/edrb8vrxeTPrLY/Jc2Z2cxX82GxmPzezl8zsgJn9u3J7\nVcck4kdVx8TM6s3sV2b2fNmP/1xuv9TMni7HzcNmFsntDODuVf0HIItSWavLANQCeB7ANdX2o+zL\nMQBrVmC/7wfwLgD7L2j7EoB7yq/vAfDFFfLj8wD+fZXHYz2Ad5VftwB4BcA11R6TiB9VHRMABqC5\n/DoH4GkAuwB8D8Any+1/A+DP38x2V+LOvhPAYXd/1Uulpx8CcMsK+LFiuPuTAN6YvH0LSoU7gSoV\n8CR+VB1373P3Z8uvx1AqjrIRVR6TiB9VxUsseZHXlQj2jQAuLLC9ksUqHcBPzWyfme1ZIR/O0+nu\nfeXXpwGEqy5Uh7vN7IXy1/xlf5y4EDPbilL9hKexgmPyBj+AKo/JchR5TX2C7np3fxeAfwvgL83s\n/SvtEFC6soOtjLH8fB3A5SitEdAH4MvV2rGZNQP4AYBPu/vohbZqjknAj6qPiS+iyCtjJYK9F8CF\ny2rQYpXLjbv3lv8fAPAjrGzlnX4zWw8A5f8HVsIJd+8vn2hFAPeiSmNiZjmUAuw77v7DcnPVxyTk\nx0qNSXnfb7rIK2Mlgv0ZANvKM4u1AD4J4JFqO2FmTWbWcv41gI8A2B/vtaw8glLhTmAFC3ieD64y\nt6IKY2JmhlINw4Pu/pULTFUdE+ZHtcdk2Yq8VmuG8Q2zjTejNNN5BMB/XCEfLkNJCXgewIFq+gHg\nuyh9Hcyj9Ox1J0pr5j0B4BCAfwLQsUJ+PADgRQAvoBRs66vgx/UofUV/AcBz5X83V3tMIn5UdUwA\nbEepiOsLKF1Y/tMF5+yvABwG8PcA6t7MdvULOiESIfUJOiGSQcEuRCIo2IVIBAW7EImgYBciERTs\nQiSCgl2IRFCwC5EI/x8cxEqbV+uOTAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Plot some images\n",
    "idx = torch.randint(0, outputs.size(0), ())\n",
    "img = images[idx,0]\n",
    "\n",
    "image = img.cpu()\n",
    "plt.imshow(image.detach().numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "7GOn8lZyOhYQ",
    "outputId": "a0c84c85-393a-4fff-ea0e-9d7341af70d0"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(42)"
      ]
     },
     "execution_count": 59,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "idx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "_upd2XRGMSMV",
    "outputId": "d6026523-f17c-43e1-fb4f-40a27905254f"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([32, 32])"
      ]
     },
     "execution_count": 60,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "image.size()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "bLX6Yjkz_y2t"
   },
   "outputs": [],
   "source": [
    "# Visualize feature maps\n",
    "activation = {}\n",
    "def get_activation(name):\n",
    "    def hook(model, input, output):\n",
    "        activation[name] = output.detach()\n",
    "    return hook\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "7KOUPjMRhhup",
    "outputId": "f41c3fc6-4bf1-4dc6-8628-a0929282492c"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<torch.utils.hooks.RemovableHandle at 0x7f4aa690fa20>"
      ]
     },
     "execution_count": 62,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.conv1.register_forward_hook(get_activation('conv1'))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 312
    },
    "colab_type": "code",
    "id": "4i4bBf71B7E9",
    "outputId": "783c9dd5-4f3c-48db-f2d9-1db1b4150873"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 0.98, 'Conv1 Feature Maps')"
      ]
     },
     "execution_count": 63,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAALMAAAEVCAYAAACrJtS0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nzBZRg/8PReTCa4vYdbMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = CNN()\n",
    "#model.CNN[0].register_forward_hook(get_activation('conv1'))\n",
    "model.conv1.register_forward_hook(get_activation('conv1'))\n",
    "x = torch.randn(1, 3, 32, 32)\n",
    "output = model(x)\n",
    "#print(output.shape)\n",
    "\n",
    "\n",
    "act = activation['conv1'].squeeze()\n",
    "num_plot = 2\n",
    "fig, axarr = plt.subplots(min(act.size(0), num_plot))\n",
    "for idx in range(min(act.size(0), num_plot)):\n",
    "    axarr[idx].imshow(act[idx])\n",
    "fig.suptitle('Conv1 Feature Maps', fontsize=16)    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "zR9iJlJj7qkN",
    "outputId": "60cadb59-326a-4f4f-bfd0-b238e0c3a0b9"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([32, 32, 32])"
      ]
     },
     "execution_count": 64,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "act.size()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "x5I--kGXFDtx"
   },
   "outputs": [],
   "source": [
    "model.conv3.register_forward_hook(get_activation('conv3'))\n",
    "x = torch.randn(1, 3, 32, 32)\n",
    "#img = img.reshape(1,1,32,32)\n",
    "output = model(x)\n",
    "#print(output.shape)\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 312
    },
    "colab_type": "code",
    "id": "tG7UrMbDF4RX",
    "outputId": "58e861da-2550-4d37-bc46-8f118a612928"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 0.98, 'Conv3 Feature Maps')"
      ]
     },
     "execution_count": 66,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAALMAAAEVCAYAAACrJtS0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjIsIGh0\ndHA6Ly9tYXRwbG90bGliLm9yZy8li6FKAAAdb0lEQVR4nO2deZxV1ZXvv4saKCiKoQoEBASUIZJE\nURE1rUZbxSG2Eo1GE6O8qGnz2k+/tHZaknw68ZN0EtImRhONrzExEH0OaBzoGKNAokDUFkwQZ0HG\nYihmqoqhxvX+OKfite7aNSDcOvfU+n4+93PvXXeffdY593f32WfdvdcWVcVx0kCPrnbAcQ4WLmYn\nNbiYndTgYnZSg4vZSQ0uZic1dFjMInKKiMwRkY0iUi8i20VknohcIyIFh9LJDvg2Q0SWi8guEdkn\nIu+IyLdFpHcHtp0lIhp4FB4if78mIpcciro/CiIyLePYxxmffzrj87O7wse26NCXJSJfA24H/gjc\nAqwFBgBTgHuAXcBTh8jHjtAX+DXwLlAHfAr4FnACcHEHtt8KXNTaqKqNB9HHTL4GLAYeP0T1f1Rq\ngC8B/97Kfk38WVnOPeoIqtrmAzgdaAZ+Fvj8KOCY9urJ9QP4IaDAwHbKzQIqc+zbGuCBg1xnz4NQ\nx7T4nM0CVgOS8VkvoJqo0VDg7K7+jls/OtLNuAXYAfxb4Mfwvqoub3kvIpNFZL6I1IrIHhFZICKT\nM7eJL+2VInKciCwSkb0iskJEbsgoc2J8OctqMUXkFyKyVUSK2vB7e/x8UFpXETlWROaKyM64K/Nn\nETmtVZkTReSx+Nj2ici7IvIDEemVUWYNMBL4YsYle1b82az489b7fl5Ens94f0a83SUicq+IbAWq\nOuNrO9wf+3hqhu2zRN3S3xr+tXvcGcexWEQuFpE3RKQu7hJe3qrcOBF5QkS2iMh+EVknIo+21+1r\nU8xxX/hM4DlV3d/m4UfljwFeIOqCTAOuJuoCvCAix7Yq3hd4EHiAqCuwBLhHRM4EUNUlRN2Gq1rt\noxj4PPCwqja0+qxQRPrE/bmbgPtUdVd7fmdsm/nokfHZ8cCLQDlwPXAp0Y9lvoickFHNEcAy4Abg\nPOBO4MtErVkLnwU2A88Cp8SP73XER4OfA0LUJZjWSV/bYi2wMK63hauBJ4Bao3xHjruFMcDPgJ8A\nlwArgYdbvveYp4FhwFeBc4HpRN3Hthvfdi47g4kuKT/s4GXqMaL+c/8MW1+ilv3xVpd2Bc7MvEwS\nnfSZGbZvAfuAfhm2qfG2k1vt+xOxveUxGyjogM+zWm3X8viPjDILgLeB4gxbQWx7MlCvEN2TXEXU\nTator5sR+7LGsD8PPJ/x/ozYxyeMsp321ehmjCES406gBBhKdIU7J2PfZjejneN+Pt725Fa+vQMs\nit8PjMtcdCi6GZ3hdOB3ma2hqlYDc4FPtyq7V1X/lFGuDniP6FfewgNEIr8sw/Yl4F1VfaVVfSuB\nE4lO9jeJWsDfdNDvLfG2mY9fAMSXyk8DjwLNLS030Zc2Pz5m4rJ9ReRHIvI+UUvSQHTJFmBsB33p\nDE9kvumMrx3gUaJz/w/AF4muJgusgp087vWq+nLLG1Vtivc1Ob4abgdWATNE5HoR6fB5ay+asZ2o\nZRzZwfrKgU2GfTNR1yOTnUa5OqKWAABVXSsiLZe7X4pIf+AzGJfluBu0NH77gohsAn4tIj/PPHkB\nGlR1aeCzcqLW49/JvrsHQER6qGoz0WX1bODbRJfdPcBk4O7M4zqItD7XnfG1TVS1RkSeJDr3o4D/\np6rNImIV78xxV5FNFVAMDFLVKhE5B7iV6Ca+QkRWA7ep6j1t+dymmFW1Mb7xOEdEesatZ1vsAIYY\n9iHY4u0I9wP3ishIov5TMVGL3R4t4hwDtCfmtthFdLm8m0BLH3/JJUR9/1tV9c6Wz0Tkk53Y136i\n42tNBR/c0H5o1wfiayf8+Q1R/7UHcKVV4ACOe3DAVk8UIkVVVwFXS/TLORa4EfiFiKxR1WdCznYk\nzjyDqK/zn8D/MQ5mNFAWRzReAC4QkTJVrYk/LyO6VD3fgX1ZPArcRXSpO5+ob7W2A9u1dGveP8D9\nAqCqe0RkEdFJ/UsbYuhJ1Co2tLJPM8rWEYW6WrMWGCwig1R1K4CIHAWMJ7qpO1i+dpR5wBxgl6q+\nGSjTmeMGGCEiJ7dcLeMgw2XAK6391agTvUxEbgKuJbovOnAxq+rCuLLbRWQC0U3KOqJuw1nAdcAX\ngOVEl/8LgQUi8iOiluMWoDfw3fb2Fdh/tYg8BfwT0Y3I9ZmfxxGUHxOJfhXRyT2d6If3jKq+dCD7\nbcVNRHf3z4rIr4gu7wOB44luMqer6m4ReRm4Oe7ibCO6iRpm1PcWcJqIXEjUBdumqmviY/ge8ICI\n3B7v4xtxXQfN145WFPdnzRY5o0xnjhuiLsUjIvIdopb4q8C4+Lnl+7wTeIToPqiA6IfRSPSnXZvO\ndOhB9K/ao0Qnp4GoS/Ec0V1rj4xyJxHdbNQS9Z0WkB15mIXxRwWt7toz7J8h+mF8KLKhH0RcHiQK\n8u8juhwvIRJ/u38khHwxyh0NPEx0s1gHVBLd2F6QUWYUUctRE5e7K8P3MzLKfQxYBOyNP5uV8dlU\n4I34WF4j+pf1Q+eF9iMK7foa2G5aXO+YNspk7bsTx/080T+fF8XHWEcUfv18RpnDiCJR78XnZwfR\nFf/c9r4jiStwnENOfP9VqKqntlf2QPBRc05qcDE7qcG7GU5q8JbZSQ0uZic1uJid1OBidlKDi9lJ\nDS5mJzW4mJ3U4GJ2UoOL2UkNLmYnNbiYndTgYnZSwyHJpXYgiEhLvoUC4JeqOqOt8sVFpVpS0j/L\nXldhTrhE6m07gAYy5RXuC/gamIzUHKinqNbOQ9PUK3z6m3ra9pCvRdX2gLGmnvZxF+2uD+5be9p+\nSaN94NX7N29T1UHBCnNEIsQczwO7mygvQyWwRETmqupboW1KSvpz4vH/lGVfeZV9SL0qw8mP6gbY\nX9LAZQEh7LHL76uwL3SDF1tzUaH66NYT1j9g11hbtfX9bdGOmGeLc/coa34sDH52XXDf9aMPM+1F\n2/eY9mff/EFH5mQecpLSzZgMrFTVVapaTzTlpyMJDx3nbyRFzMOA9RnvKzEmRIrIV0RkqYgsrW+w\nWwmn+5IUMXcIVZ2pqpNUdVJxUWlXu+MkjKSIeQMwIuP98NjmOB0mETeARKkBxsYJZTYAVxDl4gjS\na8Q+Pn7H61n2VQsnmeX3D24K1jVksX2j13z1VtO+cVtf037EkB2mfX2/w0172bpwjpZxF6ww7csr\n7XQUzdPthFEVPezjXjcgnHFNAjPpyt8OyCWUHibHJELMcRqwG4nSvBYQpaJNyCly8oVEiBlAVX8P\n/L6r/XDyl6T0mR3nI+NidlKDi9lJDYnpM3eW3XW9eGblhCz7+5//v2b5E1693LQDbOptRydYV27b\nC+0oROW27LEiALO/epdpv/rFa4M+bV49wrSX9rWXltnxtB3lqLdd4pNT3w3ue+c3jjDt9f3aWg+p\n6/GW2UkNLmYnNbiYndTgYnZSg4vZSQ0uZic15G1orriwkVEDswf2LAwsirxnX2AeElCw0z4NZavs\n33pBvT0SZ9w0e2Gr/738i6b9V6fMDvr0v/5oh+1qd/Y27b2ttauAPutsX5e8eWRw3/Mf+Klpv/TH\n5vLpicFbZic1uJid1OBidlKDi9lJDS5mJzXkbTSjvrGQdTuy8058u2mqWb7XwrJgXfsC6Ut2fdxO\n3kIveyrSS6+PtcsHBiZd+99fCfokA+w8GJPHrDHtVXOOMu0bT7W/4nFjNgX3fe5j/2ra+wS3SAbe\nMjupwcXspAYXs5MaXMxOanAxO6khMdEMEVkD1ABNQKOq2tlcYrSxB3VbsscpVPe0owB7h4bXCG+y\nE2XSd4V9er711UdM+20z7Lw1xZdvM+0b11YEfRo8qNq076qzB2FM+sFS0z62V5Vp/8+/Tgnue8rp\ny0z7kjeOC26TBBIj5pgzVdX+5h2nHbyb4aSGJIlZgedE5FURMf9NyExp21Rbm2P3nKSTpG7Gqaq6\nQUQOA+aJyDuqujCzgKrOBGYC9Bw5ItwJdroliWmZVXVD/LwFeIIom77jdJhEtMwiUgr0UNWa+PUU\n4LttbTO2XxWPXXh7ln3WjlPM8qsqBgbreueRj5n2UZfaM0e+89pFpl2G2qlxt79nD/4oHR7uKn1x\n5BLTfvebp5v2bQ/ZiVu+/K2fmPZ7Su16ABoCKw1V28M/EkMixAwMBp4QEYh8elBV/9C1Ljn5RiLE\nrKqrgGO72g8nv0lMn9lxPiouZic1uJid1JCIPvOBUNPckz/uGZdlf/L9Y8zyXzl6cbCuL/3zi6b9\nG298tlM+DfmfOtO+/ca9pr3H/PAKrXesOd/eptGOmPSYao8CuHHlFaa9ZkUg1y3w/J4S015xvL1g\nUVLwltlJDS5mJzW4mJ3U4GJ2UoOL2UkNeRvN2FzTjx8t/EyWvew9+5B+vfiCYF1Trn7J/mCRHW34\nl+ueNO3zbz3atM8ZOde0n7T65qBPcpgdGblkwl9N++PP2WNSTj7LnoGypWp4cN8Vx+yyt1lkL5uc\nFLxldlKDi9lJDS5mJzW4mJ3U4GJ2UoOL2UkNeRuak0ahZHO2+3uO22eWv+qYV4J1zb3306Z9z0l2\nXT95/GLTfukFfzbtp//X1+39XmdPaQL44UZ7oNGcZSeY9p6j7SlYr26z1+Dee4x9bADfGz3ftN/1\nQ3v98fAq3LnFW2YnNbiYndTgYnZSg4vZSQ0uZic1iGruslyJyH3AhcAWVf1EbCsHHgFGAWuAy1V1\nZ3t1lY0fosf/4qos+8ZX7MEwDX3tRXIAJh+/wrT/ZdF4015aaU9d2jvEPpf1g+yFfvq8XxT06eRL\nXzPtNw2eZ9pXNNhJbu7beJppH1u2Jbjv9fvsAVavLrbPx6p/u/nV9lIQ54Jct8yzgPNa2aYDC1R1\nLLAgfu84nSanYo4TIbZevf1ioGVF9NmAvfaZ47RDEvrMg1W1ZVG6zUSpukwyU9o27LJnPDvdlySI\n+W9o1IEPduJVdaaqTlLVSUX9s5eAcLo3SRBzlYgMBYifw3cmjtMGSRibMRe4BpgRPz/VkY3q6wtZ\nuyH7Dl4H2pGDgurwoQ4psRfDufqCP5n2v+yyxzts+dmRtv1z9hSo2o+FI0nPv2Ans5nfb4Jpr1hq\nH99/ffNO014k4ejOje9eadoHvmb7uypYU27JacssIg8BLwHjRaRSRK4lEvE5IrICODt+7zidJqct\ns6raP3k4K5d+OOkkCX1mxzkouJid1OBidlJDEqIZB0aTIDuzxzZceNqrZvFFG8Kry8x92Z69cdTH\nNpr2XoUNpn3rRLtt+Ndj7Zkbb+0NJ1VZGPB3z+vlpn3ctHdM+2ULbzDtx4zaENz3lKFvm/ZFD64O\nbpMEvGV2UoOL2UkNLmYnNbiYndTgYnZSQ95GM6SomcLDs4eBhvJE1L4bXgxn9PH2nf3aLXbkoKyP\nnXNi4KQq0/5Q5YmmXe84LOjTzs/ZYyeOO22laX/pjTGm/a6z7jft//y7acF9r33PjqTIDYGxJPc8\nFqwrl3jL7KQGF7OTGlzMTmpwMTupwcXspAYXs5Ma8jc0J0pJz+wBP/WN9iH1GR/OK7NqnR0iO2OC\nnay1rHC/aX/67U+Y9vIBe0x7/ZhwEpj+5a0zMkRMGfiWaV89pMK0T3/9EtNedLjtE8D+nWWmvWSb\nnfwmKXjL7KQGF7OTGlzMTmpwMTupwcXspIacRjMCKW1vBa4HtsbFvqmqv2+vLt1XQN1r2YOHqkfU\nm+VHPha+E6+faEcVKkf2N+11gYjJ1Al2GtrHl9rTsnqdZiefAajfU2Lab19uZ2WYONweLLX01bGm\nfcCRdrQEoLp3H9NeMzp36Y8PhCSktAX4qapOjB/tCtlxLJKQ0tZxDgpJ6TPfKCLLReQ+EQkOPM5M\nadu0Jxz0d7onSRDzPcBRwERgExBc6TEzpW1BaWmu/HPyhC4Xs6pWqWqTqjYD9wKTu9onJz/J6QI9\nACIyCvhdRjRjaEvmfBH5F+AkVb2iA/VsBdbGbwcC2w6Jw8kmKcc9UlUHdbUTuV5t6iHgDKIvoQr4\nTvx+IlHG/DXAP2YsC9HRepcmYbWjXNNdjztEElLa/iqXPjjppcv7zI5zsEiLmGd2tQNdRHc9bpOc\n3wA6zqEiLS2z47iYnfSQ12IWkfNE5F0RWSkiqV1zO/6bf4uIvJFhKxeReSKyIn4O5x/rJuStmEWk\nALgbOB+YAFwpIvYiefnPLLJHG04HFqjqWGBB/L5bk7diJvrbe6WqrlLVeuBhokXlU0dgtOHFwOz4\n9Wxgak6dSiD5LOZhwPqM95WxrbswOOOf0s3A4K50Jgnks5idGI3iq90+xprPYt4AZCZjHh7bugtV\nIjIUosFawJYu9qfLyWcxLwHGishoESkGriBaVL67MBe4Jn59DfBUF/qSCPL6H0ARuQC4AygA7lPV\n73exS4eEwGjDJ4E5wBFEQ2EvV9VuPSUtr8XsOJnkczfDcT6Ei9lJDS5mJzUkJj+ziJwH3El0M/dL\nVZ3RVvk+A4q0Ylh21p9BBfYi7Svr+gbralL7N93UbNsretppDmobi017oxaY9qHFu4M+FUmTaa+s\nt4dg9C6wMznta7KzNfUvzF52rgXFzv60o96eEV+7ompbEuYAJkLMGeMsziH6J2+JiMxVVTuzNlAx\nrIRbHstOe3VDfzvUfNEKK5FSRE19T9O+a28v0/6lo14x7S/vGm3at++3RXDLqGeCPg0rtFN33bLG\nTh4+sX+laX+reohpv+gwO5UYwP5m+wfwcGA9w0Xn/Hit+UGOSUo3o9uMs3AOHUkRc4fGWWRmNKrd\naXcnnO5LUsTcITIzGvUZEF4PxOmeJEXM3X2chXMQSMQ/gCJSCLwHnEUk4iXAF1T1zdA2JYeP0FHX\n3ZRlP+yvdvejZHP47v29L9v5iCtG2itUVS+zV3a6bupzdv177JuwUKQB4O3t9gpYtcvtfTf2sr/H\n5gH2+SjYEd534R47mjFomb04/Yu//fqrSUhGk4hohqo2isiNwLN8MM4iKGTHsUiEmAHiJOOeaNw5\nYJLSZ3acj4yL2UkNLmYnNSSmz9xZisvqGXHmuuwPzrTLr35xhP0BoIFxEPsWDzTtFWvsu/r7d55r\n2hvsYAl15XY9AH3fD7Qzh9lRix5D7PW8x95mRzN2HGOPIwGoubDWtFeOCGzz22BVOcVbZic1uJid\n1OBidlKDi9lJDS5mJzW4mJ3UkLehufqmAtbvzF6offTXa8zyRxxhh64ANpxqzyhpCkSiht+4wrS/\nPXe8aW8os0Nw2jMcmtt9nB1SGzzfHiC0f0tv015pRwsZ8YddwX3X9c8+rwDlG+0QphEg7RK8ZXZS\ng4vZSQ0uZic1uJid1OBidlJD3kYzpLaAwhezE7vUjbRDEO9f3sahFtmRg9L37cjBmnvHmfbezfYg\nIGmy24yiGnt6EsBJ1y837fO2HGfah//RTgKz/mo7ArGqT7/gvstW28ex6cLAjPg5wapyirfMTmpw\nMTupwcXspAYXs5MaXMxOakhMNENE1gA1QBPQ2F5SkeZeSu2x2eMtyt+208cePf2dYF3F/22Pzdg2\nws7e2fM2O63spk/Z2UTHnL3KtK+fc2TQp2de/4Rp71FsRxqqR9qRl0FP21/x9mOCu6Zktz1mpMe2\n8FSrJJAYMcecqarbutoJJz/xboaTGpIkZgWeE5FXReQrVoHMlLZNNXb2eqf7kqRuxqmqukFEDgPm\nicg78QLof0NVZwIzAXqOHt71GR+dRJGYlllVN8TPW4AniLLpO06HSUTLLCKlQA9VrYlfTwG+29Y2\nxcWNjB6Wfa+47oyshPsANF5izwIB6P2sHQmoG2A3/kV/Z4+p6BEYurDj5yNNuw4NusSE/7Dvg9fc\nZkdYei22M81sCvg6Zk64m7bpU3Zdw+fbB7g6WFNuSYSYgcHAEyICkU8PquofutYlJ99IhJhVdRVw\nbFf74eQ3iekzO85HxcXspAYXs5MaEtFnPhCadhax87HsyEXvnvbde+9l9pgNgIIGezaGBiaCbDzP\nLn/kyC2mvbLUjrAMebkx6FPtx+0FekZXrDftGwIzSg6fbc8o2XhLeN+F8237hjMDcknI4h3eMjup\nwcXspAYXs5MaXMxOanAxO6khb6MZzcVQawx5aC4KzJJoCP9uS6vsSMCm0wLL7g7ZbdrXvna4aS8a\nby94U/3JwGAOoOIOO6vn1l+NMu0Nw0I5OOxj6/VUOG/Gjin20szjvr/PtNvzaHKPt8xOanAxO6nB\nxeykBhezkxpczE5qcDE7qSFvQ3Na3EyzsehO2Z/thC67jrdTvgI0rLCTm5S/YU+b2tFYYdp7brPb\nhuE3v27aV8w+PuhT9T+EEq7YPo0/bq1pX1M4yrQfvji8YFHpb+xpZKsvs8OFvBGsKqd4y+ykBhez\nkxpczE5qcDE7qSGnYhaR+0Rki4i8kWErF5F5IrIifrZTbDpOO+Q6mjELuAv4TYZtOrBAVWeIyPT4\n/S3tVdSjh9K7NPuOvLnIjmYU7A4fan0g2cve4fagpRHP2uW3nGC3DSt+fpJpL+tXHfRpz5ZAetw5\n9qClt3sPN+29Jtr72FCYvbhRC/sH21OqircnOyNaTlvmOHfcjlbmi4HZ8evZwNRc+uSkhyT0mQer\n6qb49Wai7EaO02mSIOa/oapK6F8BWqW0rbbH3DrdlySIuUpEhgLEz/Z8faKUtqo6SVUnFfQN/Bvl\ndFuSIOa5wDXx62uAp7rQFyePyWk0Q0QeAs4ABopIJfAdYAYwR0SuBdYCl3ekruamHtTuym6dB07Z\napav2Rq+e2+qt3/TUybagw5WHj3ItDcEpk31CywaVPJSeOpSb7V7W7vGBxYNqrKPoXS4PSalaXM4\nMlFYa8ti/8BkRzNyKmZVvTLw0Vm59MNJJ0noZjjOQcHF7KQGF7OTGlzMTmrI25kmPQqaKe2XnZSk\n5F57nFLRleGZFX1etCMEy/9kr0yx7Vg74UrZOtu+e7w9xmNX33ASmJ6ldhSicZ3ta9HIGtNes9ce\n41E/wfYJoKTKjr4M+R97G08C4zgHGRezkxpczE5qcDE7qcHF7KSGvI1mlBQ2cPSgqix78832oLuN\nL48N1rX/XHs2RvnP7NwVtSPs2Szl79aZ9r1D7Xqa60K5MaA+MD6ieI/d/tRX2lGOslV2eSkP7prm\ngFu10+xUvjweriuXeMvspAYXs5MaXMxOanAxO6nBxeykBhezkxryNjS3r7aEZYvGZdmvvGChWf6v\nBWOCdfV+2p5Sde3Mh037998637Rvmmy3DX2ftmNd2sbZr/57e2BUXZE9cKj/cruymqPswUGl68Lt\nWK/t9jaN65KdbMpbZic1uJid1OBidlKDi9lJDUlIaXuriGwQkWXx44Jc+uSkhySktAX4qar+uDMV\naZHSODR7atGDz5xuli/ZHVpbGgo+Zw9O+sYLnzPtUmKvR021vbBN38vsxDTNGvapcJ8dtbjylD+b\n9vt7hdLm2utdV5eWBff9hb/7k2n/9dsn2xv8MlhVTklCSlvHOSgkpc98o4gsj7shyQ5mOoklCWK+\nBzgKmAhsAn4SKvihlLY1e3Lln5MndLmYVbVKVZtUtRm4F5jcRtkPUtqW2YPRne5Ll4u5JTdzzGdJ\nzHqfTr4hGkidekh2lpHSFqgiSml7BlEXQ4E1wD9mLAvRVl1biVLgEte37aA7nHySctwjVdXO85tD\ncirmQ4WILFXVSV3tR67prscdosu7GY5zsHAxO6khLWKe2dUOdBHd9bhNUtFndhxIT8vsOPktZhE5\nT0TeFZGV8brbqSQw2rBcROaJyIr4udsPA8hbMYtIAXA3cD4wAbhSRCZ0rVeHjFnAea1s04EFqjoW\nWBC/79bkrZiJ/vZeqaqrVLUeeJhoUfnUERhteDEwO349G5iaU6cSSD6LeRiwPuN9ZWzrLgzO+Kd0\nMzC4K51JAvksZidGo5BUtw9L5bOYNwAjMt4Pj23dhaqWQVrxsz1dphuRz2JeAowVkdEiUgxcQbSo\nfHdhLnBN/Poa4Kku9CUR5PWfJvHk1zuAAuA+Vf1+F7t0SAiMNnwSmAMcQTR68HJV7dZT0vJazI6T\nST53MxznQ7iYndTgYnZSg4vZSQ0uZic1uJid1OBidlKDi9lJDf8fjil0U+V3xrcAAAAASUVORK5C\nYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "act1 = activation['conv3'].squeeze()\n",
    "num_plot = 2\n",
    "fig, axarr = plt.subplots(min(act1.size(0), num_plot))\n",
    "for idx in range(min(act.size(0), num_plot)):\n",
    "    axarr[idx].imshow(act1[idx])\n",
    "fig.suptitle('Conv3 Feature Maps', fontsize=16)    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "t7lLrx91qBwn"
   },
   "outputs": [],
   "source": [
    "model.conv4.register_forward_hook(get_activation('conv4'))\n",
    "x = torch.randn(1, 3, 32, 32)\n",
    "#img = img.reshape(1,1,32,32)\n",
    "output = model(x)\n",
    "#print(output.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 312
    },
    "colab_type": "code",
    "id": "qAPlCupRpm_N",
    "outputId": "6f9f8cb3-0ead-4d5d-9f27-7461741d87a6"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 0.98, 'Conv4 Feature Maps')"
      ]
     },
     "execution_count": 70,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAALMAAAEVCAYAAACrJtS0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n4GJ2UoOL2UkNLmYnNbiYndTwf8N6iv198R0AAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "act2 = activation['conv4'].squeeze()\n",
    "num_plot = 2\n",
    "fig, axarr = plt.subplots(min(act1.size(0), num_plot))\n",
    "for idx in range(min(act.size(0), num_plot)):\n",
    "    axarr[idx].imshow(act1[idx])\n",
    "fig.suptitle('Conv4 Feature Maps', fontsize=16)    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "JPTQuHMaAHiC"
   },
   "outputs": [],
   "source": [
    "from torchvision import utils\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "\n",
    "def vistensor(tensor, ch=0, nrow=20, padding=1): \n",
    "    \n",
    "    n,c,w,h = tensor.shape\n",
    "    if c != 3:\n",
    "      tensor = tensor.view(n*c,-1,w,h )\n",
    "      tensor = tensor[:,ch,:,:].unsqueeze(dim=1)\n",
    "        \n",
    "    rows = np.min((tensor.shape[0]//nrow + 1, 64 ) )    \n",
    "    grid = utils.make_grid(tensor, nrow=nrow, normalize=True, padding=padding)\n",
    "    plt.figure(figsize=(nrow,rows) )\n",
    "    plt.imshow(grid.cpu().numpy().transpose((1, 2, 0)))\n",
    "\n",
    "\n",
    "\n",
    "def plot_img(tensor, num_cols=10):\n",
    "\tnum_kernels = tensor.shape[0]\n",
    "\tnum_rows = 1+ num_kernels // num_cols\n",
    "\tfig = plt.figure(figsize=(num_cols,num_rows))\n",
    "\tfor i in range(num_kernels):\n",
    "\t\tax1 = fig.add_subplot(num_rows,num_cols,i+1)\n",
    "\t\tax1.imshow(tensor[i], cmap='gray')\n",
    "\t\tax1.axis('off')\n",
    "\t\tax1.set_xticklabels([])\n",
    "\t\tax1.set_yticklabels([])\n",
    "\n",
    "\tplt.subplots_adjust(wspace=0.1, hspace=0.1)\n",
    "\treturn plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 600
    },
    "colab_type": "code",
    "id": "TvbPUqdNBgWN",
    "outputId": "38d74b9c-27e6-4182-b7b3-68fda6431664"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "first Kernel\n",
      "torch.Size([32, 3, 3, 3])\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA+AAAAB7CAYAAAD0bupgAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 1440x144 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Second Kernel\n",
      "torch.Size([64, 32, 3, 3])\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA+AAAAB7CAYAAAD0bupgAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nagAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 1440x144 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Third Kernel\n",
      "torch.Size([128, 64, 3, 3])\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABBoAAADnCAYAAACwnzWLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
      "text/plain": [
       "<Figure size 1440x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize Kernels\n",
    "\n",
    "optim = 'adam'\n",
    "lr = 0.0001\n",
    "drop_p = 0.5\n",
    "filters = model.modules()\n",
    "model_layers = [i for i in model.children()]\n",
    "first_layer = model_layers[0]\n",
    "second_layer = model_layers[3]\n",
    "third_layer = model_layers[5]\n",
    "\n",
    "first_kernels = first_layer.weight.data.clone()\n",
    "print(\"first Kernel\")\n",
    "print(first_kernels.shape)\n",
    "vistensor(first_kernels)\n",
    "plt.axis('off')\n",
    "plt.ioff()\n",
    "plt.show()\n",
    "\n",
    "\n",
    "second_kernels = second_layer.weight.data.clone()\n",
    "print(\"Second Kernel\")\n",
    "print(second_kernels.shape)\n",
    "vistensor(second_kernels[0, ...].reshape(1,32,3,3))\n",
    "plt.axis('off')\n",
    "plt.ioff()\n",
    "plt.show()\n",
    "\n",
    "third_kernels = third_layer.weight.data.clone()\n",
    "print(\"Third Kernel\")\n",
    "print(third_kernels.shape)\n",
    "vistensor(third_kernels[0, ...].reshape(1,64,3,3))\n",
    "plt.axis('off')\n",
    "plt.ioff()\n",
    "plt.show()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "b5s7GB9HH7xz",
    "outputId": "f2211d60-35be-477c-f648-0483d3e99168"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))"
      ]
     },
     "execution_count": 46,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_layers[5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "G7vjQ1rSXiZ6"
   },
   "source": [
    "Q)Which one is working better in CIFAR-10, MaxPooling or AvgPooling?\n",
    "\n",
    "Average Pooling works better."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "BOclnpmtXbel"
   },
   "source": [
    "Q) Write the relation between ConvLayer and FC:\n",
    "\n",
    "\n",
    "The only difference between FC and CONV layers is that the neurons in the CONV layer are connected only to a local region in the input, and that many of the neurons in a CONV volume share parameters. However, the neurons in both layers still compute dot products, so their functional form is identical."
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "CUDA_ASS4.ipynb",
   "provenance": []
  },
  "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": 1
}