Chess_Piece_Detection_(Using_Transfer_Learning).ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/DataMinati/Super-Notebooks-100/blob/main/Chess_Piece_Detection_(Using_Transfer_Learning).ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "d058b04c",
      "metadata": {
        "papermill": {
          "duration": 0.013083,
          "end_time": "2022-01-27T01:30:49.802051",
          "exception": false,
          "start_time": "2022-01-27T01:30:49.788968",
          "status": "completed"
        },
        "tags": [],
        "id": "d058b04c"
      },
      "source": [
        "# Import Necessary Libraries"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Data Source : https://www.kaggle.com/taranmarley/squeezenet-transfer-learning/data"
      ],
      "metadata": {
        "id": "XlcMWZfePULd"
      },
      "id": "XlcMWZfePULd"
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "882fb8b1",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2022-01-27T01:30:49.831079Z",
          "iopub.status.busy": "2022-01-27T01:30:49.829944Z",
          "iopub.status.idle": "2022-01-27T01:30:51.070480Z",
          "shell.execute_reply": "2022-01-27T01:30:51.069587Z",
          "shell.execute_reply.started": "2022-01-26T16:37:21.173217Z"
        },
        "papermill": {
          "duration": 1.256542,
          "end_time": "2022-01-27T01:30:51.070677",
          "exception": false,
          "start_time": "2022-01-27T01:30:49.814135",
          "status": "completed"
        },
        "tags": [],
        "id": "882fb8b1"
      },
      "outputs": [],
      "source": [
        "import numpy as np\n",
        "import pandas as pd \n",
        "import torch\n",
        "from torch import optim\n",
        "import torch.nn as nn\n",
        "from PIL import Image\n",
        "import os"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "4edc53c9",
      "metadata": {
        "papermill": {
          "duration": 0.01053,
          "end_time": "2022-01-27T01:30:51.092227",
          "exception": false,
          "start_time": "2022-01-27T01:30:51.081697",
          "status": "completed"
        },
        "tags": [],
        "id": "4edc53c9"
      },
      "source": [
        "# Create dataframe with class ids and filenames"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "2920d119",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2022-01-27T01:30:51.121280Z",
          "iopub.status.busy": "2022-01-27T01:30:51.120470Z",
          "iopub.status.idle": "2022-01-27T01:30:51.435606Z",
          "shell.execute_reply": "2022-01-27T01:30:51.436117Z",
          "shell.execute_reply.started": "2022-01-26T16:37:22.620745Z"
        },
        "papermill": {
          "duration": 0.333228,
          "end_time": "2022-01-27T01:30:51.436297",
          "exception": false,
          "start_time": "2022-01-27T01:30:51.103069",
          "status": "completed"
        },
        "tags": [],
        "id": "2920d119",
        "outputId": "9b394883-efcd-4021-b6ce-4ef8c768c175"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "                                             fileloc      filename  \\\n",
            "0  ../input/chess-pieces-detection-images-dataset...  00000030.jpg   \n",
            "1  ../input/chess-pieces-detection-images-dataset...  00000042.jpg   \n",
            "2  ../input/chess-pieces-detection-images-dataset...  00000027.jpg   \n",
            "3  ../input/chess-pieces-detection-images-dataset...  00000127.jpg   \n",
            "4  ../input/chess-pieces-detection-images-dataset...  00000113.jpg   \n",
            "\n",
            "        classid  int_class_id  \n",
            "0  KnightImages             0  \n",
            "1  KnightImages             0  \n",
            "2  KnightImages             0  \n",
            "3  KnightImages             0  \n",
            "4  KnightImages             0  \n"
          ]
        }
      ],
      "source": [
        "class_ids = [\"KnightImages\", \"PawnImages\", \"Queenimages\", \"RookImage\", \"bishopImages\"]\n",
        "\n",
        "y = []\n",
        "X = []\n",
        "filenames = []\n",
        "for class_id in class_ids:\n",
        "    for file in os.listdir(os.path.join(\"../input/chess-pieces-detection-images-dataset\", class_id)):\n",
        "        X.append(os.path.join(\"../input/chess-pieces-detection-images-dataset\", class_id, file))\n",
        "        y.append(class_id)\n",
        "        filenames.append(file)\n",
        "        \n",
        "df = pd.DataFrame(list(zip(X, filenames, y)), columns =['fileloc', 'filename', 'classid'])\n",
        "df[\"int_class_id\"] = df[\"classid\"].astype(\"category\").cat.codes\n",
        "print(df.head())"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "5e715d56",
      "metadata": {
        "papermill": {
          "duration": 0.010668,
          "end_time": "2022-01-27T01:30:51.458336",
          "exception": false,
          "start_time": "2022-01-27T01:30:51.447668",
          "status": "completed"
        },
        "tags": [],
        "id": "5e715d56"
      },
      "source": [
        "# Create the Custom Dataset Class\n",
        "\n",
        "We need this to be able to load the image and label into the model we will create. So we will create a custom dataset to handle this"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "8f45187a",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2022-01-27T01:30:51.486849Z",
          "iopub.status.busy": "2022-01-27T01:30:51.485907Z",
          "iopub.status.idle": "2022-01-27T01:30:51.493328Z",
          "shell.execute_reply": "2022-01-27T01:30:51.493734Z",
          "shell.execute_reply.started": "2022-01-26T16:37:22.742794Z"
        },
        "papermill": {
          "duration": 0.021958,
          "end_time": "2022-01-27T01:30:51.493939",
          "exception": false,
          "start_time": "2022-01-27T01:30:51.471981",
          "status": "completed"
        },
        "tags": [],
        "id": "8f45187a"
      },
      "outputs": [],
      "source": [
        "from torch.utils.data import Dataset\n",
        "\n",
        "class CustomDataset(Dataset):\n",
        "  def __init__(self, X, y, BatchSize, transform):\n",
        "    super().__init__()\n",
        "    self.BatchSize = BatchSize\n",
        "    self.y = y\n",
        "    self.X = X\n",
        "    self.transform = transform\n",
        "    \n",
        "  def num_of_batches(self):\n",
        "    \"\"\"\n",
        "    Detect the total number of batches\n",
        "    \"\"\"\n",
        "    return math.floor(len(self.list_IDs) / self.BatchSize)\n",
        "\n",
        "  def __getitem__(self,idx):\n",
        "    class_id = self.y[idx]\n",
        "    img = Image.open(self.X[idx])\n",
        "    img = img.convert(\"RGBA\").convert(\"RGB\")\n",
        "    img = self.transform(img)\n",
        "    return img, torch.tensor(int(class_id))\n",
        "\n",
        "  def __len__(self):\n",
        "    return len(self.X)"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "3468d8e2",
      "metadata": {
        "papermill": {
          "duration": 0.010588,
          "end_time": "2022-01-27T01:30:51.515321",
          "exception": false,
          "start_time": "2022-01-27T01:30:51.504733",
          "status": "completed"
        },
        "tags": [],
        "id": "3468d8e2"
      },
      "source": [
        "# Instantiate the Datasets\n",
        "\n",
        "We will form them into torch dataloaders to make the data easier to work with. We are also going to put in a minor amount of image augmentation in the train dataset."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "4897e2b9",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2022-01-27T01:30:51.541033Z",
          "iopub.status.busy": "2022-01-27T01:30:51.540085Z",
          "iopub.status.idle": "2022-01-27T01:30:52.767171Z",
          "shell.execute_reply": "2022-01-27T01:30:52.767656Z",
          "shell.execute_reply.started": "2022-01-26T16:37:22.760104Z"
        },
        "papermill": {
          "duration": 1.241639,
          "end_time": "2022-01-27T01:30:52.767842",
          "exception": false,
          "start_time": "2022-01-27T01:30:51.526203",
          "status": "completed"
        },
        "tags": [],
        "id": "4897e2b9"
      },
      "outputs": [],
      "source": [
        "from sklearn.model_selection import train_test_split\n",
        "from torch.utils.data import DataLoader\n",
        "from torchvision import transforms\n",
        "\n",
        "# Shuffle dataframe\n",
        "df = df.sample(frac=1)\n",
        "\n",
        "X = df.iloc[:,0]\n",
        "y = df.iloc[:,3]\n",
        "\n",
        "transform = transforms.Compose([\n",
        "                transforms.Resize([256,256]),\n",
        "                transforms.RandomRotation(20, fill=256),\n",
        "                transforms.ColorJitter(brightness=0.1, contrast=0.1, saturation=0.05, hue=0.05),\n",
        "                transforms.ToTensor(),\n",
        "                transforms.RandomAffine(degrees=0, translate=(0.025, 0.025), fill=256),\n",
        "                transforms.Normalize([0.5], [0.5])\n",
        "            ])\n",
        "\n",
        "test_transform = transforms.Compose([\n",
        "                transforms.Resize([256,256]),\n",
        "                transforms.ToTensor(),\n",
        "                transforms.Normalize((0.5,), (0.5,)),\n",
        "            ])\n",
        "\n",
        "train_ratio = 0.90\n",
        "validation_ratio = 0.05\n",
        "test_ratio = 0.05\n",
        "\n",
        "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=1 - train_ratio, stratify = y, random_state = 0)\n",
        "X_val, X_test, y_val, y_test = train_test_split(X_test, y_test, test_size=test_ratio/(test_ratio + validation_ratio), random_state = 0)\n",
        "\n",
        "dataset_stages = ['train', 'val', 'test']\n",
        "\n",
        "batch_size = 32\n",
        "image_datasets = {'train' : CustomDataset(X_train.values, y_train.values, batch_size, transform), 'val' : CustomDataset(X_val.values, y_val.values, batch_size, test_transform), 'test' : CustomDataset(X_test.values, y_test.values, batch_size, test_transform)}\n",
        "dataloaders = {x: DataLoader(image_datasets[x], batch_size=image_datasets[x].BatchSize,\n",
        "                                            shuffle=True, num_workers=0)\n",
        "            for x in dataset_stages}\n",
        "\n",
        "dataset_sizes = {x: len(image_datasets[x]) for x in ['train', 'val', 'test']}"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "4d1b3b44",
      "metadata": {
        "papermill": {
          "duration": 0.010539,
          "end_time": "2022-01-27T01:30:52.792071",
          "exception": false,
          "start_time": "2022-01-27T01:30:52.781532",
          "status": "completed"
        },
        "tags": [],
        "id": "4d1b3b44"
      },
      "source": [
        "# Test image from dataset"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "797a9b2a",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2022-01-27T01:30:52.818955Z",
          "iopub.status.busy": "2022-01-27T01:30:52.818305Z",
          "iopub.status.idle": "2022-01-27T01:30:53.004010Z",
          "shell.execute_reply": "2022-01-27T01:30:53.004569Z",
          "shell.execute_reply.started": "2022-01-26T16:37:23.781999Z"
        },
        "papermill": {
          "duration": 0.201704,
          "end_time": "2022-01-27T01:30:53.004769",
          "exception": false,
          "start_time": "2022-01-27T01:30:52.803065",
          "status": "completed"
        },
        "tags": [],
        "id": "797a9b2a",
        "outputId": "340987ee-f86f-4fe0-8589-c7996cf7d624"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<PIL.Image.Image image mode=RGB size=256x256 at 0x7F8A887EFCD0>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "nparray = image_datasets['train'][2][0].cpu().numpy() \n",
        "image = transforms.ToPILImage()(image_datasets['train'][2][0].cpu()).convert(\"RGB\")\n",
        "display(image)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "6ed26a71",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2022-01-27T01:30:53.034142Z",
          "iopub.status.busy": "2022-01-27T01:30:53.033171Z",
          "iopub.status.idle": "2022-01-27T01:30:53.038266Z",
          "shell.execute_reply": "2022-01-27T01:30:53.038813Z",
          "shell.execute_reply.started": "2022-01-26T16:37:23.960492Z"
        },
        "papermill": {
          "duration": 0.021303,
          "end_time": "2022-01-27T01:30:53.038977",
          "exception": false,
          "start_time": "2022-01-27T01:30:53.017674",
          "status": "completed"
        },
        "tags": [],
        "id": "6ed26a71",
        "outputId": "77f9a081-5c24-459d-bdae-eaf28fc51546"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "' for i in range(0,10000):\\n    print(i)\\n    nparray = image_datasets[\\'train\\'][i][0].cpu().numpy() \\n    image = transforms.ToPILImage()(image_datasets[\\'train\\'][i][0].cpu()).convert(\"RGB\")\\n'"
            ]
          },
          "execution_count": 6,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "\"\"\" for i in range(0,10000):\n",
        "    print(i)\n",
        "    nparray = image_datasets['train'][i][0].cpu().numpy() \n",
        "    image = transforms.ToPILImage()(image_datasets['train'][i][0].cpu()).convert(\"RGB\")\n",
        "\"\"\""
      ]
    },
    {
      "cell_type": "markdown",
      "id": "4f972d32",
      "metadata": {
        "papermill": {
          "duration": 0.013216,
          "end_time": "2022-01-27T01:30:53.065080",
          "exception": false,
          "start_time": "2022-01-27T01:30:53.051864",
          "status": "completed"
        },
        "tags": [],
        "id": "4f972d32"
      },
      "source": [
        "# Create a Training Function"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "7672e347",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2022-01-27T01:30:53.105499Z",
          "iopub.status.busy": "2022-01-27T01:30:53.104534Z",
          "iopub.status.idle": "2022-01-27T01:30:53.107520Z",
          "shell.execute_reply": "2022-01-27T01:30:53.106924Z",
          "shell.execute_reply.started": "2022-01-26T16:37:23.969382Z"
        },
        "papermill": {
          "duration": 0.029565,
          "end_time": "2022-01-27T01:30:53.107655",
          "exception": false,
          "start_time": "2022-01-27T01:30:53.078090",
          "status": "completed"
        },
        "tags": [],
        "id": "7672e347"
      },
      "outputs": [],
      "source": [
        "import time\n",
        "\n",
        "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
        "\n",
        "def train_model(model, criterion, optimizer, scheduler, num_epochs=10):\n",
        "    since = time.time()\n",
        "    best_acc = 0.0\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        print('Epoch {}/{}'.format(epoch, num_epochs - 1))\n",
        "        print('-' * 10)\n",
        "        # Each epoch has a training and validation phase\n",
        "        for phase in ['train', 'val']:\n",
        "            if phase == 'train':\n",
        "                model.train()  # Set model to training mode\n",
        "            else:\n",
        "                model.eval()   # Set model to evaluate mode\n",
        "\n",
        "            running_loss = 0.0\n",
        "            running_corrects = 0\n",
        "            num_batches = 0\n",
        "            outputs = None\n",
        "            # Iterate over data.\n",
        "            for inputs, labels in dataloaders[phase]:\n",
        "                # Loading Bar\n",
        "                if (phase == 'train'):\n",
        "                    num_batches += 1\n",
        "                    percentage_complete = ((num_batches * batch_size) / (dataset_sizes[phase])) * 100\n",
        "                    percentage_complete = np.clip(percentage_complete, 0, 100)\n",
        "                    print(\"{:0.2f}\".format(percentage_complete), \"% complete\", end=\"\\r\")\n",
        "\n",
        "                inputs = inputs.to(device)\n",
        "                labels = labels.to(device)\n",
        "\n",
        "                # zero the parameter gradients\n",
        "                optimizer.zero_grad()\n",
        "                \n",
        "                # forward\n",
        "                # track history if only in train\n",
        "                with torch.set_grad_enabled(phase == 'train'):\n",
        "                    outputs = model(inputs)\n",
        "                    loss = criterion(outputs.float(), labels)\n",
        "\n",
        "                    # backward + optimize only if in training phase\n",
        "                    if phase == 'train':\n",
        "                        loss.backward()\n",
        "                        # TODO: try removal\n",
        "                        torch.nn.utils.clip_grad_norm_(model.parameters(), 1)\n",
        "                        optimizer.step()\n",
        "\n",
        "                # statistics\n",
        "                running_loss += loss.item() * inputs.size(0)\n",
        "                \n",
        "                predicted = torch.max(outputs.data, 1)[1] \n",
        "                running_correct = (predicted == labels).sum()\n",
        "                running_corrects += running_correct\n",
        "            if phase == 'train':\n",
        "                scheduler.step()\n",
        "\n",
        "            epoch_loss = running_loss / dataset_sizes[phase]\n",
        "            \n",
        "            epoch_acc = running_corrects / dataset_sizes[phase]\n",
        "            #epoch_acc = sum(epoch_acc) / len(epoch_acc)\n",
        "\n",
        "            print('{} Loss: {:.4f} Acc: {:.4f}'.format(\n",
        "                phase, epoch_loss, epoch_acc.item()))\n",
        "\n",
        "    time_elapsed = time.time() - since\n",
        "    print('Training complete in {:.0f}m {:.0f}s'.format(\n",
        "        time_elapsed // 60, time_elapsed % 60))\n",
        "    return model"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "5179b72a",
      "metadata": {
        "papermill": {
          "duration": 0.012702,
          "end_time": "2022-01-27T01:30:53.133546",
          "exception": false,
          "start_time": "2022-01-27T01:30:53.120844",
          "status": "completed"
        },
        "tags": [],
        "id": "5179b72a"
      },
      "source": [
        "# Load up EfficientNet"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "4c43933a",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2022-01-27T01:30:53.164429Z",
          "iopub.status.busy": "2022-01-27T01:30:53.163819Z",
          "iopub.status.idle": "2022-01-27T01:30:53.165337Z",
          "shell.execute_reply": "2022-01-27T01:30:53.165729Z",
          "shell.execute_reply.started": "2022-01-26T16:37:24.028499Z"
        },
        "papermill": {
          "duration": 0.018964,
          "end_time": "2022-01-27T01:30:53.165920",
          "exception": false,
          "start_time": "2022-01-27T01:30:53.146956",
          "status": "completed"
        },
        "tags": [],
        "id": "4c43933a"
      },
      "outputs": [],
      "source": [
        "from torchvision import models\n",
        "from torch.optim import lr_scheduler"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "6ed89b61",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2022-01-27T01:30:53.199332Z",
          "iopub.status.busy": "2022-01-27T01:30:53.198315Z",
          "iopub.status.idle": "2022-01-27T01:30:54.764284Z",
          "shell.execute_reply": "2022-01-27T01:30:54.764817Z",
          "shell.execute_reply.started": "2022-01-26T16:37:24.040898Z"
        },
        "papermill": {
          "duration": 1.585791,
          "end_time": "2022-01-27T01:30:54.764990",
          "exception": false,
          "start_time": "2022-01-27T01:30:53.179199",
          "status": "completed"
        },
        "tags": [],
        "colab": {
          "referenced_widgets": [
            "f38627409f784822be0336b1a52bc75e"
          ]
        },
        "id": "6ed89b61",
        "outputId": "08dcedd6-00b0-4b5b-be91-ec350e862a56"
      },
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Downloading: \"https://download.pytorch.org/models/squeezenet1_1-b8a52dc0.pth\" to /root/.cache/torch/hub/checkpoints/squeezenet1_1-b8a52dc0.pth\n"
          ]
        },
        {
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "f38627409f784822be0336b1a52bc75e",
              "version_major": 2,
              "version_minor": 0
            },
            "text/plain": [
              "  0%|          | 0.00/4.73M [00:00<?, ?B/s]"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "\n",
        "model_ft = models.squeezenet1_1(pretrained=True)\n",
        "model_ft.classifier._modules[\"1\"] = nn.Conv2d(512, 5, kernel_size=(1, 1))\n",
        "model_ft.num_classes = 5\n",
        "for param in model_ft.parameters():\n",
        "    param.requires_grad = False\n",
        "for param in model_ft.classifier.parameters():\n",
        "    param.requires_grad = True"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "8471b9f7",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2022-01-27T01:30:54.796501Z",
          "iopub.status.busy": "2022-01-27T01:30:54.795527Z",
          "iopub.status.idle": "2022-01-27T01:42:30.231972Z",
          "shell.execute_reply": "2022-01-27T01:42:30.232444Z",
          "shell.execute_reply.started": "2022-01-26T16:37:24.324049Z"
        },
        "papermill": {
          "duration": 695.454013,
          "end_time": "2022-01-27T01:42:30.232858",
          "exception": false,
          "start_time": "2022-01-27T01:30:54.778845",
          "status": "completed"
        },
        "tags": [],
        "id": "8471b9f7",
        "outputId": "2582a39f-e757-47d8-ea43-d133e621dd07"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 0/19\n",
            "----------\n",
            "train Loss: 41.5943 Acc: 0.1931\n",
            "val Loss: 1.5862 Acc: 0.2647\n",
            "Epoch 1/19\n",
            "----------\n",
            "train Loss: 2.7518 Acc: 0.2162\n",
            "val Loss: 1.5049 Acc: 0.3529\n",
            "Epoch 2/19\n",
            "----------\n",
            "train Loss: 1.5611 Acc: 0.2690\n",
            "val Loss: 1.6659 Acc: 0.3235\n",
            "Epoch 3/19\n",
            "----------\n",
            "train Loss: 1.5326 Acc: 0.3036\n",
            "val Loss: 1.4445 Acc: 0.4412\n",
            "Epoch 4/19\n",
            "----------\n",
            "train Loss: 1.4503 Acc: 0.3498\n",
            "val Loss: 1.3500 Acc: 0.4706\n",
            "Epoch 5/19\n",
            "----------\n",
            "train Loss: 1.3804 Acc: 0.3960\n",
            "val Loss: 1.3192 Acc: 0.5294\n",
            "Epoch 6/19\n",
            "----------\n",
            "train Loss: 1.3482 Acc: 0.4373\n",
            "val Loss: 1.0904 Acc: 0.6765\n",
            "Epoch 7/19\n",
            "----------\n",
            "train Loss: 1.2335 Acc: 0.4818\n",
            "val Loss: 1.0232 Acc: 0.7941\n",
            "Epoch 8/19\n",
            "----------\n",
            "train Loss: 1.2148 Acc: 0.5017\n",
            "val Loss: 0.9298 Acc: 0.7647\n",
            "Epoch 9/19\n",
            "----------\n",
            "train Loss: 1.1897 Acc: 0.5281\n",
            "val Loss: 0.8764 Acc: 0.7647\n",
            "Epoch 10/19\n",
            "----------\n",
            "train Loss: 1.1884 Acc: 0.5050\n",
            "val Loss: 0.8663 Acc: 0.7647\n",
            "Epoch 11/19\n",
            "----------\n",
            "train Loss: 1.1708 Acc: 0.5462\n",
            "val Loss: 0.8612 Acc: 0.7941\n",
            "Epoch 12/19\n",
            "----------\n",
            "train Loss: 1.1487 Acc: 0.5479\n",
            "val Loss: 0.8150 Acc: 0.7059\n",
            "Epoch 13/19\n",
            "----------\n",
            "train Loss: 1.1243 Acc: 0.5644\n",
            "val Loss: 0.7851 Acc: 0.7647\n",
            "Epoch 14/19\n",
            "----------\n",
            "train Loss: 1.0157 Acc: 0.6403\n",
            "val Loss: 0.7835 Acc: 0.7647\n",
            "Epoch 15/19\n",
            "----------\n",
            "train Loss: 1.0827 Acc: 0.5759\n",
            "val Loss: 0.7828 Acc: 0.7353\n",
            "Epoch 16/19\n",
            "----------\n",
            "train Loss: 1.0959 Acc: 0.6007\n",
            "val Loss: 0.7834 Acc: 0.7353\n",
            "Epoch 17/19\n",
            "----------\n",
            "train Loss: 1.1042 Acc: 0.5842\n",
            "val Loss: 0.7848 Acc: 0.7353\n",
            "Epoch 18/19\n",
            "----------\n",
            "train Loss: 1.0936 Acc: 0.5677\n",
            "val Loss: 0.7829 Acc: 0.7353\n",
            "Epoch 19/19\n",
            "----------\n",
            "train Loss: 1.0533 Acc: 0.5941\n",
            "val Loss: 0.7862 Acc: 0.7353\n",
            "Training complete in 11m 35s\n"
          ]
        }
      ],
      "source": [
        "criterion = nn.CrossEntropyLoss()\n",
        "\n",
        "optimizer_ft = optim.Adam(model_ft.parameters(), lr=0.01)\n",
        "\n",
        "exp_lr_scheduler = lr_scheduler.StepLR(optimizer_ft, step_size=7, gamma=0.1)\n",
        "\n",
        "model_ft = train_model(model_ft.to(device), criterion, optimizer_ft, exp_lr_scheduler, 20)"
      ]
    },
    {
      "cell_type": "markdown",
      "id": "50496c9d",
      "metadata": {
        "papermill": {
          "duration": 0.106801,
          "end_time": "2022-01-27T01:42:30.449810",
          "exception": false,
          "start_time": "2022-01-27T01:42:30.343009",
          "status": "completed"
        },
        "tags": [],
        "id": "50496c9d"
      },
      "source": [
        "# Run on Test Set"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "id": "6c76ac52",
      "metadata": {
        "execution": {
          "iopub.execute_input": "2022-01-27T01:42:30.675885Z",
          "iopub.status.busy": "2022-01-27T01:42:30.675253Z",
          "iopub.status.idle": "2022-01-27T01:42:32.295084Z",
          "shell.execute_reply": "2022-01-27T01:42:32.294536Z",
          "shell.execute_reply.started": "2022-01-26T16:43:23.954886Z"
        },
        "papermill": {
          "duration": 1.736126,
          "end_time": "2022-01-27T01:42:32.295249",
          "exception": false,
          "start_time": "2022-01-27T01:42:30.559123",
          "status": "completed"
        },
        "tags": [],
        "id": "6c76ac52",
        "outputId": "007d31cb-75eb-4718-d458-19bc7d4a7d56"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Accuracy: 0.529411792755127\n"
          ]
        }
      ],
      "source": [
        "from sklearn.metrics import accuracy_score \n",
        "\n",
        "accuracy_scores = []\n",
        "\n",
        "running_corrects = 0\n",
        "outputs = None\n",
        "for inputs, labels in dataloaders['test']:\n",
        "    model_ft.eval()\n",
        "    \n",
        "    inputs = inputs.to(device)\n",
        "    labels = labels.to(device)\n",
        "\n",
        "    outputs = model_ft(inputs)\n",
        "    \n",
        "    predicted = torch.max(outputs.data, 1)[1] \n",
        "    running_correct = (predicted == labels).sum()\n",
        "    running_corrects += running_correct\n",
        "\n",
        "accuracy = running_corrects / dataset_sizes['test']\n",
        "print(\"Accuracy: \" + str(accuracy.item()))"
      ]
    }
  ],
  "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": "3.7.12"
    },
    "papermill": {
      "default_parameters": {},
      "duration": 712.424434,
      "end_time": "2022-01-27T01:42:33.316892",
      "environment_variables": {},
      "exception": null,
      "input_path": "__notebook__.ipynb",
      "output_path": "__notebook__.ipynb",
      "parameters": {},
      "start_time": "2022-01-27T01:30:40.892458",
      "version": "2.3.3"
    },
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {}
    },
    "colab": {
      "name": "Chess Piece Detection (Using Transfer Learning)",
      "provenance": [],
      "include_colab_link": true
    }
  },
  "nbformat": 4,
  "nbformat_minor": 5
}