QuickDraw_Recognition_small_data.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "QuickDraw_Recognition-small_data.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": []
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.1"
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "metadata": {
        "id": "7aeEoQtEhIvq",
        "colab_type": "code",
        "outputId": "7d9f4959-999d-401d-cf5f-11ba111c6349",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 122
        }
      },
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/gdrive')"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3Aietf%3Awg%3Aoauth%3A2.0%3Aoob&scope=email%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdocs.test%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive.photos.readonly%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fpeopleapi.readonly&response_type=code\n",
            "\n",
            "Enter your authorization code:\n",
            "··········\n",
            "Mounted at /content/gdrive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "n1PCzGsghPCD",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "dataset_path = '/content/gdrive/My Drive/quickdraw_small/'"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "YhLUrbsy_L7c",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "%reload_ext autoreload\n",
        "%autoreload 2\n",
        "%matplotlib inline"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "0CCtvCAv8gts",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "from PIL import Image\n",
        "import os\n",
        "\n",
        "import torch\n",
        "from torch import nn, optim\n",
        "from torchvision import transforms, datasets"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "6K1FCqzahHMP",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "from my_ml_lib import MetricTools, PlotTools"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "rwWVAKjnTNNv",
        "outputId": "7c8f9fb5-699d-4437-825c-62a6907adf53",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
        "device"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "device(type='cuda', index=0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "CiiEfPGB_d7K",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "data_dirs = {\n",
        "    'train': dataset_path+'train/',\n",
        "    'valid': dataset_path+'valid/',\n",
        "    'test': dataset_path+'test/',\n",
        "}"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "QVkxEEhRhHMe",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "for d in ['train', 'valid', 'test']:\n",
        "    os.rmdir(data_dirs[d] + '.ipynb_checkpoints') "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "scrolled": true,
        "id": "HBfay6TMhHMg",
        "colab_type": "code",
        "outputId": "14f4fdf0-979b-4fd2-99f3-fff82423a642",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 374
        }
      },
      "cell_type": "code",
      "source": [
        "files = os.listdir(data_dirs['train'])\n",
        "files = files[:10]\n",
        "idx_to_class = sorted([f.split('_')[-1].split('.')[0] for f in files])\n",
        "\n",
        "class_to_idx = {idx_to_class[i]: i for i in range(len(idx_to_class))}\n",
        "\n",
        "len(idx_to_class), class_to_idx, idx_to_class"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(10,\n",
              " {'airplane': 0,\n",
              "  'alarm clock': 1,\n",
              "  'banana': 2,\n",
              "  'baseball bat': 3,\n",
              "  'bicycle': 4,\n",
              "  'car': 5,\n",
              "  'crown': 6,\n",
              "  'fish': 7,\n",
              "  'headphones': 8,\n",
              "  'pants': 9},\n",
              " ['airplane',\n",
              "  'alarm clock',\n",
              "  'banana',\n",
              "  'baseball bat',\n",
              "  'bicycle',\n",
              "  'car',\n",
              "  'crown',\n",
              "  'fish',\n",
              "  'headphones',\n",
              "  'pants'])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 9
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "K83Yo1D2_3Yn",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "data_transforms = {\n",
        "    'train':transforms.Compose([\n",
        "            transforms.RandomRotation(10),\n",
        "            transforms.ToTensor(),\n",
        "#             transforms.Normalize([0.155], [0.316])\n",
        "        ]),\n",
        "    'valid':transforms.Compose([\n",
        "            transforms.ToTensor(),\n",
        "#             transforms.Normalize([0.155], [0.316])\n",
        "        ]),\n",
        "    'test':transforms.Compose([\n",
        "            transforms.ToTensor(),\n",
        "#             transforms.Normalize([0.155], [0.316])\n",
        "        ])\n",
        "}"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "DqIRZhg4hHMo",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "dataset = {}\n",
        "# mean, std = None, None\n",
        "\n",
        "for d in ['train', 'valid', 'test']:\n",
        "    data_x = []\n",
        "    data_y = []\n",
        "    for path, _, _temp in os.walk(data_dirs[d]):\n",
        "        for f in files:\n",
        "            c = f.split('_')[-1].split('.')[0] # get class name from file name\n",
        "            x = np.load(path + f).reshape(-1, 28, 28) / 255\n",
        "            y = np.ones((len(x), 1), dtype=np.int64) * class_to_idx[c]\n",
        "            \n",
        "            data_x.extend(x)\n",
        "            data_y.extend(y)\n",
        "#     if d == 'train':\n",
        "#         mean = np.mean(data_x)\n",
        "#         std = np.std(data_x)\n",
        "    dataset[d] = torch.utils.data.TensorDataset(torch.stack([data_transforms[d](Image.fromarray(np.uint8(i*255))) for i in data_x]), \n",
        "                                                torch.stack([torch.Tensor(j) for j in data_y]))\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "_omZ50EuhHMr",
        "colab_type": "code",
        "outputId": "b06e61f2-4c01-40f3-e8a1-3becfe4f5325",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "[len(dataset[d]) for d in ['train', 'valid', 'test']]"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[84000, 36000, 30000]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "metadata": {
        "id": "jgKksvPyhHMw",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "bs = 128 # batch size"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "LYTWu6yyhHMy",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "dataloaders = {\n",
        "    d: torch.utils.data.DataLoader(dataset[d], batch_size=bs, shuffle=True) for d in ['train', \n",
        "                                                                                      'valid', \n",
        "                                                                                      'test']\n",
        "}"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "nLIBfFgZMYcC",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "invTrans = transforms.Compose([ transforms.Normalize([0.],[1/0.316]),\n",
        "                                transforms.Normalize([-0.155],[1.]),\n",
        "                               ])\n",
        "\n",
        "def imshow(img):\n",
        "#     img = invTrans(img)\n",
        "#     img = img.numpy()\n",
        "    plt.imshow(img.reshape(28,28), cmap='gray')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "5n_S8EUaQD3T",
        "outputId": "84d52753-56fa-4111-9f47-89251b46f58d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 213
        }
      },
      "cell_type": "code",
      "source": [
        "dataiter = iter(dataloaders['train'])\n",
        "images, labels = dataiter.next()\n",
        "\n",
        "fig = plt.figure(figsize=(8, 3))\n",
        "# display 9 images\n",
        "for cls in np.arange(10):\n",
        "    ax = fig.add_subplot(2, 5, cls+1, xticks=[], yticks=[])\n",
        "    \n",
        "    image = images.numpy()[labels.numpy().reshape(-1) == cls][0]\n",
        "    label = labels.numpy()[labels.numpy().reshape(-1) == cls][0, 0].astype(np.int64)\n",
        "    imshow(image)\n",
        "    ax.set_title(idx_to_class[label])\n",
        "plt.savefig('data_vis_small.png')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAdAAAADECAYAAAAvbXA5AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n4+TAH6CO4ziOk4P/B2H5Qx8a6qQmAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 576x216 with 10 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "J0yiBCFZQWU5",
        "outputId": "5393f160-69c0-488a-90cc-8a7e56b59dfd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 391
        }
      },
      "cell_type": "code",
      "source": [
        "# define the CNN architecture\n",
        "class Net(nn.Module):\n",
        "    def __init__(self):\n",
        "        super(Net, self).__init__()\n",
        "        \n",
        "        self.layer1 = nn.Sequential(\n",
        "            nn.Conv2d(1, 5, kernel_size=3, padding=1),\n",
        "            nn.ReLU(),\n",
        "            nn.Conv2d(5, 5, kernel_size=3, padding=1),\n",
        "            nn.ReLU(),\n",
        "            nn.MaxPool2d(2), # 14x14x5\n",
        "            nn.Conv2d(5, 8, kernel_size=3, padding=1),\n",
        "            nn.ReLU(),\n",
        "            nn.Conv2d(8, 8, kernel_size=3, padding=1),\n",
        "            nn.ReLU(),\n",
        "            nn.MaxPool2d(2), # 7x7x8\n",
        "            nn.Conv2d(8, 16, kernel_size=3, padding=1),\n",
        "            nn.ReLU() # 7x7x16\n",
        "        )\n",
        "        self.fc = nn.Sequential(\n",
        "            nn.Linear(7*7*16, 100),\n",
        "            nn.ReLU(),\n",
        "            nn.Dropout(p=0.2),\n",
        "            nn.Linear(100, 10)\n",
        "        )\n",
        "    \n",
        "    def forward(self, x):\n",
        "        out = self.layer1(x)\n",
        "        out = out.view(out.shape[0], -1)\n",
        "        out = self.fc(out)\n",
        "        return out\n",
        "\n",
        "\n",
        "model_scratch = Net()\n",
        "model_scratch.to(device)"
      ],
      "execution_count": 37,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Net(\n",
              "  (layer1): Sequential(\n",
              "    (0): Conv2d(1, 5, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
              "    (1): ReLU()\n",
              "    (2): Conv2d(5, 5, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
              "    (3): ReLU()\n",
              "    (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
              "    (5): Conv2d(5, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
              "    (6): ReLU()\n",
              "    (7): Conv2d(8, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
              "    (8): ReLU()\n",
              "    (9): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
              "    (10): Conv2d(8, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
              "    (11): ReLU()\n",
              "  )\n",
              "  (fc): Sequential(\n",
              "    (0): Linear(in_features=784, out_features=100, bias=True)\n",
              "    (1): ReLU()\n",
              "    (2): Dropout(p=0.2)\n",
              "    (3): Linear(in_features=100, out_features=10, bias=True)\n",
              "  )\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 37
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "Ro8Eu6QIUUNP",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "criterion_scratch = nn.CrossEntropyLoss()\n",
        "\n",
        "# optimizer_scratch = optim.SGD(model_scratch.parameters(), lr=0.001, momentum=0.9)\n",
        "optimizer_scratch = optim.Adam(model_scratch.parameters(), lr=0.001)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "2RoFcWVwTYvt",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def train(n_epochs, loaders, model, optimizer, criterion, device, save_path):\n",
        "    \"\"\"returns trained model\"\"\"\n",
        "    # initialize tracker for minimum validation loss\n",
        "    valid_loss_min = np.Inf \n",
        "    train_losses = []\n",
        "    val_losses = []\n",
        "    \n",
        "    for epoch in range(1, n_epochs+1):\n",
        "        # initialize variables to monitor training and validation loss\n",
        "        train_loss = 0.0\n",
        "        valid_loss = 0.0\n",
        "        \n",
        "        ###################\n",
        "        # train the model #\n",
        "        ###################\n",
        "        model.train()\n",
        "        for batch_idx, (data, target) in enumerate(loaders['train']):\n",
        "            # move to GPU or CPU\n",
        "#             print(torch.max(target,1)[0])\n",
        "            data, target = data.to(device), target.long().to(device)\n",
        "            \n",
        "            ## find the loss and update the model parameters accordingly\n",
        "            ## record the average training loss as follows\n",
        "            ## train_loss = train_loss + ((1 / (batch_idx + 1)) * (loss.data - train_loss))\n",
        "            optimizer.zero_grad()\n",
        "            output = model(data)\n",
        "            loss = criterion(output, torch.max(target,1)[0])\n",
        "            \n",
        "            loss.backward()\n",
        "            optimizer.step()\n",
        "            \n",
        "            train_loss += ((1 / (batch_idx + 1)) * (loss.data - train_loss))\n",
        "            \n",
        "        ######################    \n",
        "        # validate the model #\n",
        "        ######################\n",
        "        model.eval()\n",
        "        for batch_idx, (data, target) in enumerate(loaders['valid']):\n",
        "            # move to GPU or CPU\n",
        "            data, target = data.to(device), target.long().to(device)\n",
        "            \n",
        "            ## update the average validation loss\n",
        "            output = model(data)\n",
        "            loss = criterion(output, torch.max(target,1)[0])\n",
        "            \n",
        "            valid_loss += ((1 / (batch_idx + 1)) * (loss.data - valid_loss))\n",
        "            \n",
        "        # print training/validation statistics \n",
        "        print('Epoch: {} \\tTraining Loss: {:.6f} \\tValidation Loss: {:.6f}'.format(\n",
        "            epoch, \n",
        "            train_loss,\n",
        "            valid_loss\n",
        "            ))\n",
        "        \n",
        "        train_losses.append(train_loss)\n",
        "        val_losses.append(valid_loss)\n",
        "        \n",
        "        if valid_loss < valid_loss_min:\n",
        "            print('Saving model..')\n",
        "            valid_loss_min = valid_loss\n",
        "            torch.save(model.state_dict(), save_path)\n",
        "    \n",
        "    # return trained model\n",
        "    return model, train_losses, val_losses"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "ToF8RLJtU_LP",
        "outputId": "455a9f9e-627b-4ccb-bd7c-19165712c301",
        "scrolled": true,
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 646
        }
      },
      "cell_type": "code",
      "source": [
        "# train the model\n",
        "model_scratch, train_losses, val_losses = train(25, dataloaders, model_scratch, optimizer_scratch, \n",
        "                                                criterion_scratch, device, 'model_scratch-25epochs-small.pt')"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch: 1 \tTraining Loss: 0.888627 \tValidation Loss: 0.442564\n",
            "Saving model..\n",
            "Epoch: 2 \tTraining Loss: 0.480836 \tValidation Loss: 0.331642\n",
            "Saving model..\n",
            "Epoch: 3 \tTraining Loss: 0.383856 \tValidation Loss: 0.278752\n",
            "Saving model..\n",
            "Epoch: 4 \tTraining Loss: 0.326465 \tValidation Loss: 0.253496\n",
            "Saving model..\n",
            "Epoch: 5 \tTraining Loss: 0.288928 \tValidation Loss: 0.236648\n",
            "Saving model..\n",
            "Epoch: 6 \tTraining Loss: 0.263565 \tValidation Loss: 0.228176\n",
            "Saving model..\n",
            "Epoch: 7 \tTraining Loss: 0.246785 \tValidation Loss: 0.215645\n",
            "Saving model..\n",
            "Epoch: 8 \tTraining Loss: 0.235600 \tValidation Loss: 0.212701\n",
            "Saving model..\n",
            "Epoch: 9 \tTraining Loss: 0.223981 \tValidation Loss: 0.205885\n",
            "Saving model..\n",
            "Epoch: 10 \tTraining Loss: 0.216250 \tValidation Loss: 0.207738\n",
            "Epoch: 11 \tTraining Loss: 0.208286 \tValidation Loss: 0.200292\n",
            "Saving model..\n",
            "Epoch: 12 \tTraining Loss: 0.198608 \tValidation Loss: 0.205331\n",
            "Epoch: 13 \tTraining Loss: 0.194282 \tValidation Loss: 0.205973\n",
            "Epoch: 14 \tTraining Loss: 0.188328 \tValidation Loss: 0.199760\n",
            "Saving model..\n",
            "Epoch: 15 \tTraining Loss: 0.182125 \tValidation Loss: 0.198810\n",
            "Saving model..\n",
            "Epoch: 16 \tTraining Loss: 0.178107 \tValidation Loss: 0.199221\n",
            "Epoch: 17 \tTraining Loss: 0.172410 \tValidation Loss: 0.206759\n",
            "Epoch: 18 \tTraining Loss: 0.168530 \tValidation Loss: 0.210828\n",
            "Epoch: 19 \tTraining Loss: 0.166675 \tValidation Loss: 0.205812\n",
            "Epoch: 20 \tTraining Loss: 0.159625 \tValidation Loss: 0.208815\n",
            "Epoch: 21 \tTraining Loss: 0.157984 \tValidation Loss: 0.214351\n",
            "Epoch: 22 \tTraining Loss: 0.153650 \tValidation Loss: 0.213153\n",
            "Epoch: 23 \tTraining Loss: 0.149516 \tValidation Loss: 0.216772\n",
            "Epoch: 24 \tTraining Loss: 0.146618 \tValidation Loss: 0.204432\n",
            "Epoch: 25 \tTraining Loss: 0.143968 \tValidation Loss: 0.221465\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "tplg8-eUhHNQ",
        "colab_type": "code",
        "outputId": "100804e9-2f9f-41b8-9cf4-cef2edfc5d43",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        }
      },
      "cell_type": "code",
      "source": [
        "plt.plot(train_losses, label='Training loss')\n",
        "plt.plot(val_losses, label='Validation loss')\n",
        "plt.xticks([i for i in range(0, len(train_losses), 5)].append(len(train_losses)))\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Loss')\n",
        "plt.legend()\n",
        "_ = plt.ylim()\n",
        "# plt.title('Training Validation Loss with CNN from scratch')\n",
        "# plt.savefig('Training Validation Loss with CNN from scratch-25epochs-small.png')\n",
        "plt.show()"
      ],
      "execution_count": 41,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEKCAYAAAD9xUlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nY0yIhYIxxpgQCwVjjDEh/x+C8udtqTrxCwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "e_whUDl2VuOi",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# load the model that got the best validation accuracy\n",
        "model_scratch.load_state_dict(torch.load('model_scratch-25epochs-small.pt'))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "lxCSa_lhhHNY",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def mapk(target, output, k=3):\n",
        "    map_sum = 0\n",
        "    output = torch.topk(output, k)[1]\n",
        "    for i, t in enumerate(target):\n",
        "        idx = (output[i] == t).nonzero().cpu().numpy()\n",
        "        if len(idx) == 0:\n",
        "            continue\n",
        "            \n",
        "        idx = idx[0][0]\n",
        "        map_sum += 1 / (idx + 1)\n",
        "        \n",
        "    return map_sum"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "b66HUWouVydb",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def test(loaders, model, criterion, device):\n",
        "    sigmoid = lambda x: 1 / (1 + np.exp(-x))\n",
        "    \n",
        "    # monitor test loss and accuracy\n",
        "    test_loss = 0.\n",
        "    correct = 0.\n",
        "    total = 0.\n",
        "    y = None\n",
        "    y_hat = None\n",
        "    total_mapk = 0\n",
        "    \n",
        "    model.eval()\n",
        "    for batch_idx, (data, target) in enumerate(loaders['test']):\n",
        "        # move to GPU or CPU\n",
        "        data, target = data.to(device), target.long().to(device)\n",
        "        \n",
        "        # forward pass: compute predicted outputs by passing inputs to the model\n",
        "        output = model(data)\n",
        "        # calculate the loss\n",
        "        loss = criterion(output, target.view(-1))\n",
        "        # update average test loss \n",
        "        test_loss = test_loss + ((1 / (batch_idx + 1)) * (loss.data - test_loss))\n",
        "        # convert output probabilities to predicted class\n",
        "        pred = output.data.max(1, keepdim=True)[1]\n",
        "        \n",
        "        if y is None:\n",
        "            y = target.cpu().numpy()\n",
        "            y_hat = pred.data.cpu().view_as(target).numpy()\n",
        "            probs = output.data.cpu().numpy()\n",
        "        else:\n",
        "            y = np.append(y, target.cpu().numpy())\n",
        "            y_hat = np.append(y_hat, pred.data.cpu().view_as(target).numpy())\n",
        "            probs = np.vstack([probs, output.data.cpu().numpy()])\n",
        "            \n",
        "        total_mapk += mapk(target, output)\n",
        "        \n",
        "        # compare predictions to true label\n",
        "        correct += np.sum(np.squeeze(pred.eq(target.data.view_as(pred))).cpu().numpy())\n",
        "        total += data.size(0)\n",
        "            \n",
        "    print('Test Loss: {:.6f}\\n'.format(test_loss))\n",
        "\n",
        "    print('\\nTest Accuracy: %2d%% (%2d/%2d)' % (\n",
        "        100. * correct / total, correct, total))\n",
        "    \n",
        "    map3 = np.around(total_mapk / len(loaders[\"test\"].dataset) * 100, 2)\n",
        "    print(f'\\nMean Average Precision @ 3: {map3}%')\n",
        "          \n",
        "    return y, y_hat, probs"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "xjPJHHFhn-tT",
        "outputId": "ec1455e1-2149-4b89-f2c1-bdad69202c06",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 119
        }
      },
      "cell_type": "code",
      "source": [
        "# call test function    \n",
        "y, y_hat, probs = test(dataloaders, model_scratch, criterion_scratch, device)"
      ],
      "execution_count": 45,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test Loss: 0.200899\n",
            "\n",
            "\n",
            "Test Accuracy: 94% (28348/30000)\n",
            "\n",
            "Mean Average Precision @ 3: 95.99%\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "r22qqjzwboNb",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def cmc_curve(probs, y, idx_to_class):\n",
        "    counts = np.zeros(len(idx_to_class))\n",
        "    \n",
        "    indices = np.argsort(probs * -1)\n",
        "    for i in range(y.shape[0]):\n",
        "        actual = y[i]\n",
        "        rank = 1\n",
        "        for pred in indices[i]:\n",
        "            if pred == actual:\n",
        "                break\n",
        "            rank += 1\n",
        "        \n",
        "        counts[rank - 1] += 1\n",
        "    \n",
        "    ranks = np.zeros(len(idx_to_class))\n",
        "    ranks[0] = counts[0]\n",
        "    for i in range(1, len(idx_to_class)):\n",
        "        ranks[i] = ranks[i-1] + counts[i]\n",
        "\n",
        "    return ranks / y.shape[0]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "yNwc3R1pZI7O",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "cmcs = cmc_curve(probs, y, idx_to_class)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "BHARRPMdiNR2",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "cmcs_lda_3 = [0.91458333, 0.95911111, 0.97458333, 0.98288889, 0.98794444, 0.99166667,\n",
        " 0.99483333, 0.99686111, 0.99872222, 1.]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "cIAtzzSecyVT",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "95e94ed4-beae-43d7-fa8a-7cda7936ae16"
      },
      "cell_type": "code",
      "source": [
        "plt.plot([i for i in range(1,11)], cmcs_lda_3, 'bo-', label='XGB-LDA', c='r')\n",
        "plt.plot([i for i in range(1,11)], cmcs, 'bo-', label='CNN')\n",
        "plt.locator_params(axis='x', nbins=10)\n",
        "plt.xlabel('Rank')\n",
        "plt.ylabel('Identification Accuracy')\n",
        "plt.title('CMC Curve')\n",
        "plt.legend()\n",
        "plt.savefig('cmc_small.png')\n",
        "plt.show()"
      ],
      "execution_count": 68,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n/yxNmq8+l1ZnAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "_n8UDpgchHNr",
        "colab_type": "code",
        "outputId": "dad6aafe-2243-4f33-d2d0-a42f3b4d03de",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "MetricTools.accuracy(y, y_hat)"
      ],
      "execution_count": 60,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.9449333333333333"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 60
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "4Uq50WGDgFCy",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "cm = MetricTools.confusion_matrix(y, y_hat, nclasses=len(idx_to_class))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "34ioObSMqD-3",
        "outputId": "64a0129f-9086-4aad-c781-dbd63c948e20",
        "scrolled": true,
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 487
        }
      },
      "cell_type": "code",
      "source": [
        "PlotTools.confusion_matrix(cm, idx_to_class, title='',\n",
        "                           filename='Confusion Matrix with CNN from scratch-25-small', figsize=(7,7), path='./')"
      ],
      "execution_count": 62,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAesAAAHWCAYAAABXF6HSAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n+X9+XG4ngjkEMQAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 504x504 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "c5PuktzrqGHw",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "prec, rec = MetricTools.prec_recall(y, y_hat, len(idx_to_class))\n",
        "f1 = 1 /(0.5 * (1/prec + 1/rec))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "colab_type": "code",
        "id": "7VxDhXHhsqQI",
        "outputId": "cd217ca5-5b2a-4d50-aa32-a392c62372a7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "np.around([prec, rec, f1], 3)"
      ],
      "execution_count": 64,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([0.945, 0.945, 0.945])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 64
        }
      ]
    },
    {
      "metadata": {
        "id": "Kq4WsBxEhHN_",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}