03_handwritten_recognition_explore.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "cfc52381",
   "metadata": {
    "id": "cfc52381"
   },
   "source": [
    "[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/biodatlab/deep-learning-skooldio/blob/master/03_handwritten_recognition_explore.ipynb)\n",
    "\n",
    "## **Thai-digit handwritten classification with Pytorch: Student Notebook Exploration**\n",
    "\n",
    "This notebook contains exploration of the \"Thai-digit handwritten classification\" including\n",
    "- Customizing model\n",
    "    - Adding layers\n",
    "    - Adding dropout layer\n",
    "- Adding image augmentation"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3aa56f94",
   "metadata": {
    "id": "3aa56f94"
   },
   "source": [
    "## **Download and clean the dataset from the repository**\n",
    "\n",
    "- We have downloaded the data from https://github.com/kittinan/thai-handwriting-number by cloning the repository\n",
    "- Remove files that have character mismatch (as suggested by the the creator)\n",
    "- Then, we put the cleaned data at https://github.com/biodatlab/deep-learning-skooldio"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "f7deb2f6",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "f7deb2f6",
    "outputId": "e7277312-f5e6-4c5a-d3a3-905f93f1c9cb"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cloning into 'deep-learning-skooldio'...\n",
      "remote: Enumerating objects: 1990, done.\u001b[K\n",
      "remote: Counting objects: 100% (284/284), done.\u001b[K\n",
      "remote: Compressing objects: 100% (159/159), done.\u001b[K\n",
      "remote: Total 1990 (delta 153), reused 233 (delta 120), pack-reused 1706\u001b[K\n",
      "Receiving objects: 100% (1990/1990), 14.75 MiB | 29.32 MiB/s, done.\n",
      "Resolving deltas: 100% (164/164), done.\n"
     ]
    }
   ],
   "source": [
    "!git clone https://github.com/biodatlab/deep-learning-skooldio"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "29268a34",
   "metadata": {
    "id": "29268a34"
   },
   "source": [
    "After cloning the repository, check that `deep-learning-skooldio` should appear in the directory."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "2ad2a443",
   "metadata": {
    "id": "2ad2a443"
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import os.path as op\n",
    "from glob import glob\n",
    "from pathlib import Path\n",
    "import random\n",
    "from PIL import Image\n",
    "from collections import Counter\n",
    "from torch.utils.data import ConcatDataset\n",
    "from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "64ce2573",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "64ce2573",
    "outputId": "394ab2dc-f2e0-43a0-98db-5d2bfc601e53"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of samples 1750\n"
     ]
    }
   ],
   "source": [
    "directory = \"deep-learning-skooldio/\"\n",
    "paths = glob(op.join(directory, \"thai-handwritten-dataset\", \"*\", \"*\"))\n",
    "num_samples = len(paths)\n",
    "print(\"Number of samples\", num_samples)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "6aa0f5b4",
   "metadata": {
    "id": "6aa0f5b4"
   },
   "outputs": [],
   "source": [
    "from tqdm.auto import tqdm\n",
    "from sklearn.model_selection import train_test_split\n",
    "import shutil\n",
    "\n",
    "train_paths, test_paths = train_test_split(paths, test_size=0.1, random_state=42)\n",
    "for i in range(10):\n",
    "    os.makedirs(f\"data/train/{i}\", exist_ok=True)\n",
    "    os.makedirs(f\"data/validation/{i}\", exist_ok=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "fe1d237b",
   "metadata": {
    "id": "fe1d237b"
   },
   "outputs": [],
   "source": [
    "def copy_to_destination(src_paths, dst_path: str = \"data/train/\"):\n",
    "    for path in tqdm(src_paths):\n",
    "        path = Path(path)\n",
    "        parent_dir = path.parent.name\n",
    "        shutil.copy(path, op.join(dst_path, parent_dir, path.name))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "af9064ae",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 81,
     "referenced_widgets": [
      "04bff9d1c6b44aba879b1c1bd99eacd3",
      "e585b2c5a15649c1ba70dab769b0baa0",
      "95666913c37848a7995ceff186e516a1",
      "35ad23534198412db6e4233a01d34a61",
      "986f13b440ef4477b4a026ddca6e42c0",
      "546987ef32044cf0a967ee6cf5ff567b",
      "e33ce989c88d4e1da8aedc4a988e55fb",
      "c1005f423da84e93ba9fdb1788ecb9da",
      "45ab66db880c4c02a326ef3bf7c7c43c",
      "a950a090716e4dfca165a5ac2eb898d3",
      "73b451e0cc17454db355df0008c5b78e",
      "eb9c6b8de45c47dc8fac7cad28d2f6d9",
      "8d69cd19ce0141e1bb6cf76c94b3fd0f",
      "792e5f6e426b4dcc98914f0502c207c4",
      "561963521ebb40cdad9789b56981f431",
      "8aaa4f0f5ac54220b650540a0a6c4d6f",
      "83df746b0eee4a9eba9783a903db697c",
      "99b59fba156b4c1d99bb8d0c23eeba16",
      "1b225e76df10484181b6b41a6a587d00",
      "755bcd54d24e4787a6875b0a004c3536",
      "a0e3edff6bbc4871b46e8aa80479973a",
      "872049e09056436692c61b449c1b2769"
     ]
    },
    "id": "af9064ae",
    "outputId": "48e0a0bd-a99c-4c78-ade7-cff41a86820d"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "04bff9d1c6b44aba879b1c1bd99eacd3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/1575 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "eb9c6b8de45c47dc8fac7cad28d2f6d9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/175 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "copy_to_destination(train_paths, \"data/train/\")\n",
    "copy_to_destination(test_paths, \"data/validation/\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "molxWFBdMq55",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "molxWFBdMq55",
    "outputId": "1f7be324-8917-4272-d497-4fdb9e66f117"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1575, 175)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(glob(\"data/train/*/*\")), len(glob(\"data/validation/*/*\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ae55f494",
   "metadata": {
    "id": "ae55f494"
   },
   "source": [
    "## **Create a custom dataset and a dataloader**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "d0b69d68",
   "metadata": {
    "id": "d0b69d68"
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch.utils.data import Dataset, DataLoader\n",
    "from torchvision import transforms\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "2dbd4be3",
   "metadata": {
    "id": "2dbd4be3"
   },
   "outputs": [],
   "source": [
    "transform = transforms.Compose([\n",
    "    transforms.Resize((28, 28)),\n",
    "    transforms.Grayscale(),\n",
    "    transforms.ToTensor()\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "09eb4140",
   "metadata": {
    "id": "09eb4140"
   },
   "outputs": [],
   "source": [
    "class ThaiDigitDataset(Dataset):\n",
    "    def __init__(self, img_dir: str, transform=None):\n",
    "        self.img_dir = img_dir\n",
    "        self.transform = transform\n",
    "        self.img_labels = [(p, Path(p).parent.name) for p in glob(op.join(img_dir, \"*\", \"*\"))]\n",
    "\n",
    "    def __len__(self):\n",
    "        return len(self.img_labels)\n",
    "\n",
    "    def __getitem__(self, idx):\n",
    "        image, label = self.img_labels[idx]\n",
    "        label = int(label)\n",
    "        image = Image.open(image)\n",
    "        if self.transform:\n",
    "            image = 1 - self.transform(image)\n",
    "        return image, label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "f1175f91",
   "metadata": {
    "id": "f1175f91"
   },
   "outputs": [],
   "source": [
    "train_thaidigit_dataset = ThaiDigitDataset(\"data/train/\", transform=transform)\n",
    "val_thaidigit_dataset = ThaiDigitDataset(\"data/validation/\", transform=transform)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "d4f83af9",
   "metadata": {
    "id": "d4f83af9"
   },
   "outputs": [],
   "source": [
    "train_loader = DataLoader(train_thaidigit_dataset, batch_size=16, shuffle=True)\n",
    "val_loader = DataLoader(val_thaidigit_dataset, batch_size=16, shuffle=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "275542ac",
   "metadata": {
    "id": "275542ac"
   },
   "source": [
    "## **Create the model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "abbd30d0",
   "metadata": {
    "id": "abbd30d0"
   },
   "outputs": [],
   "source": [
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "\n",
    "class ThaiDigitNet(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(ThaiDigitNet, self).__init__()\n",
    "        self.fc1 = nn.Linear(28 * 28, 100)\n",
    "        self.fc2 = nn.Linear(100, 10)\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = x.view(-1, 28 * 28)\n",
    "        x = self.fc1(x)\n",
    "        x = F.relu(x)\n",
    "        x = self.fc2(x)\n",
    "        return x"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f0455ef9",
   "metadata": {
    "id": "f0455ef9"
   },
   "source": [
    "## **Train the model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "r61vWix7nfgl",
   "metadata": {
    "id": "r61vWix7nfgl"
   },
   "outputs": [],
   "source": [
    "# Create training loop function\n",
    "def train(model, n_epochs, loss_function, optimizer, train_loader, validation_loader):\n",
    "    training_logs = {\"train_loss\": [],  \"train_acc\": [], \"val_loss\": [], \"val_acc\": []}\n",
    "    print(\"-\"*80)\n",
    "    for epoch in range(1, n_epochs+1):\n",
    "        # training\n",
    "        model.train()\n",
    "        train_loss, correct = 0, 0\n",
    "        for images, labels in train_loader:\n",
    "            pred = model(images)\n",
    "            loss = loss_function(pred, labels)\n",
    "\n",
    "            optimizer.zero_grad()\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "\n",
    "            train_loss += loss.item()\n",
    "            correct += (pred.argmax(1) == labels).float().sum().item()\n",
    "        # save training logs\n",
    "        training_logs[\"train_loss\"].append(train_loss/ len(train_loader))\n",
    "        training_logs[\"train_acc\"].append(correct / len(train_loader.dataset))\n",
    "\n",
    "        # validation\n",
    "        model.eval()\n",
    "        val_loss, correct = 0, 0\n",
    "        for images, labels in validation_loader:\n",
    "            pred = model(images)\n",
    "            val_loss += loss_function(pred, labels).item()\n",
    "            correct += (pred.argmax(1) == labels).float().sum().item()\n",
    "        # save validation logs\n",
    "        training_logs[\"val_loss\"].append(val_loss/ len(val_loader))\n",
    "        training_logs[\"val_acc\"].append(correct / len(validation_loader.dataset))\n",
    "\n",
    "        print(f\"Epochs {epoch}\".ljust(10), \n",
    "              f\"train loss {training_logs['train_loss'][-1]:.5f}\",\n",
    "              f\"train acc {training_logs['train_acc'][-1]:.5f}\",\n",
    "\n",
    "              f\"val loss {training_logs['val_loss'][-1]:.5f}\",\n",
    "              f\"val acc {training_logs['val_acc'][-1]:.5f}\",\n",
    "              )\n",
    "        print(\"-\"*80)\n",
    "    return model, training_logs\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "Aa6vTE2V9GEt",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Aa6vTE2V9GEt",
    "outputId": "139b218b-4ae9-4efc-d923-0f9473bd4f72"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--------------------------------------------------------------------------------\n",
      "Epochs 1   train loss 2.29104 train acc 0.11048 val loss 2.27734 val acc 0.15429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 2   train loss 2.25231 train acc 0.21524 val loss 2.23748 val acc 0.25143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 3   train loss 2.20640 train acc 0.27111 val loss 2.18769 val acc 0.30857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 4   train loss 2.15121 train acc 0.28825 val loss 2.12486 val acc 0.39429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 5   train loss 2.08295 train acc 0.37397 val loss 2.05143 val acc 0.42857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 6   train loss 2.00475 train acc 0.40635 val loss 1.96576 val acc 0.45714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 7   train loss 1.91385 train acc 0.46159 val loss 1.87048 val acc 0.49143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 8   train loss 1.81948 train acc 0.49651 val loss 1.77095 val acc 0.54286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 9   train loss 1.72363 train acc 0.53778 val loss 1.68081 val acc 0.53714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 10  train loss 1.63414 train acc 0.54857 val loss 1.59019 val acc 0.57714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 11  train loss 1.55328 train acc 0.57397 val loss 1.51524 val acc 0.60571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 12  train loss 1.48528 train acc 0.59302 val loss 1.44810 val acc 0.60000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 13  train loss 1.42327 train acc 0.59873 val loss 1.39650 val acc 0.57714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 14  train loss 1.36851 train acc 0.60508 val loss 1.35315 val acc 0.58286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 15  train loss 1.31746 train acc 0.62730 val loss 1.30649 val acc 0.60000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 16  train loss 1.27432 train acc 0.63048 val loss 1.26904 val acc 0.61714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 17  train loss 1.24050 train acc 0.64508 val loss 1.25339 val acc 0.61714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 18  train loss 1.21003 train acc 0.65460 val loss 1.21646 val acc 0.64571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 19  train loss 1.18014 train acc 0.66095 val loss 1.19576 val acc 0.63429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 20  train loss 1.15661 train acc 0.66159 val loss 1.17262 val acc 0.64000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 21  train loss 1.13196 train acc 0.67238 val loss 1.15265 val acc 0.65143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 22  train loss 1.10651 train acc 0.68000 val loss 1.13747 val acc 0.65714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 23  train loss 1.08897 train acc 0.68444 val loss 1.12388 val acc 0.64571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 24  train loss 1.07078 train acc 0.68381 val loss 1.11822 val acc 0.63429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 25  train loss 1.05142 train acc 0.69460 val loss 1.09298 val acc 0.66286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 26  train loss 1.03380 train acc 0.69016 val loss 1.09310 val acc 0.65143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 27  train loss 1.01593 train acc 0.69841 val loss 1.08323 val acc 0.65714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 28  train loss 1.00093 train acc 0.70476 val loss 1.06988 val acc 0.66286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 29  train loss 0.98606 train acc 0.70222 val loss 1.06121 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 30  train loss 0.97434 train acc 0.69968 val loss 1.05555 val acc 0.64571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 31  train loss 0.95665 train acc 0.71429 val loss 1.05077 val acc 0.65143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 32  train loss 0.94578 train acc 0.71492 val loss 1.04309 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 33  train loss 0.93384 train acc 0.71810 val loss 1.03254 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 34  train loss 0.91756 train acc 0.72635 val loss 1.03104 val acc 0.65143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 35  train loss 0.90707 train acc 0.72698 val loss 1.02107 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 36  train loss 0.89278 train acc 0.73206 val loss 1.01270 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 37  train loss 0.88442 train acc 0.73651 val loss 1.00559 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 38  train loss 0.87303 train acc 0.74159 val loss 1.00375 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 39  train loss 0.85553 train acc 0.74921 val loss 1.00768 val acc 0.66286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 40  train loss 0.84830 train acc 0.74921 val loss 1.00497 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 41  train loss 0.84037 train acc 0.75556 val loss 1.00039 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 42  train loss 0.82599 train acc 0.75873 val loss 1.00273 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 43  train loss 0.81625 train acc 0.75619 val loss 1.00229 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 44  train loss 0.80534 train acc 0.76635 val loss 0.99853 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 45  train loss 0.79682 train acc 0.77016 val loss 0.99365 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 46  train loss 0.78486 train acc 0.77460 val loss 0.98537 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 47  train loss 0.77564 train acc 0.77968 val loss 0.98995 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 48  train loss 0.76582 train acc 0.77968 val loss 0.98492 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 49  train loss 0.75601 train acc 0.77460 val loss 0.98204 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 50  train loss 0.74864 train acc 0.78921 val loss 0.98924 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 51  train loss 0.73905 train acc 0.79111 val loss 0.97520 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 52  train loss 0.72973 train acc 0.79556 val loss 0.97621 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 53  train loss 0.71968 train acc 0.79492 val loss 0.97670 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 54  train loss 0.71363 train acc 0.80063 val loss 0.97754 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 55  train loss 0.70257 train acc 0.81079 val loss 0.96930 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 56  train loss 0.69463 train acc 0.80635 val loss 0.97330 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 57  train loss 0.68716 train acc 0.81206 val loss 0.96677 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 58  train loss 0.68121 train acc 0.81714 val loss 0.96602 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 59  train loss 0.67298 train acc 0.81651 val loss 0.96658 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 60  train loss 0.66302 train acc 0.81714 val loss 0.95786 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 61  train loss 0.65165 train acc 0.82540 val loss 0.97126 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 62  train loss 0.64475 train acc 0.82794 val loss 0.97045 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 63  train loss 0.63659 train acc 0.82730 val loss 0.95818 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 64  train loss 0.62727 train acc 0.83556 val loss 0.96544 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 65  train loss 0.62044 train acc 0.83619 val loss 0.95701 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 66  train loss 0.61607 train acc 0.83429 val loss 0.96660 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 67  train loss 0.60242 train acc 0.84190 val loss 0.96425 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 68  train loss 0.59802 train acc 0.84381 val loss 0.95748 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 69  train loss 0.59007 train acc 0.84889 val loss 0.95765 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 70  train loss 0.58275 train acc 0.84762 val loss 0.95423 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 71  train loss 0.57433 train acc 0.85524 val loss 0.95604 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 72  train loss 0.56946 train acc 0.85460 val loss 0.96735 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 73  train loss 0.56433 train acc 0.85397 val loss 0.94586 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 74  train loss 0.55488 train acc 0.85524 val loss 0.96180 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 75  train loss 0.54732 train acc 0.86349 val loss 0.95935 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 76  train loss 0.54345 train acc 0.86413 val loss 0.96277 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 77  train loss 0.53317 train acc 0.86603 val loss 0.95377 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 78  train loss 0.52563 train acc 0.87429 val loss 0.95742 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 79  train loss 0.51695 train acc 0.86984 val loss 0.95409 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 80  train loss 0.51411 train acc 0.87873 val loss 0.95369 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 81  train loss 0.50609 train acc 0.87937 val loss 0.96070 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 82  train loss 0.49940 train acc 0.88127 val loss 0.95884 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 83  train loss 0.49303 train acc 0.87873 val loss 0.95664 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 84  train loss 0.48605 train acc 0.88508 val loss 0.96264 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 85  train loss 0.48140 train acc 0.88571 val loss 0.95731 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 86  train loss 0.47547 train acc 0.88571 val loss 0.96098 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 87  train loss 0.46964 train acc 0.89333 val loss 0.96870 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 88  train loss 0.46306 train acc 0.89333 val loss 0.94662 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 89  train loss 0.45525 train acc 0.89333 val loss 0.95245 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 90  train loss 0.44981 train acc 0.89778 val loss 0.96015 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 91  train loss 0.44623 train acc 0.89905 val loss 0.95819 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 92  train loss 0.43779 train acc 0.89778 val loss 0.96148 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 93  train loss 0.43235 train acc 0.90222 val loss 0.96145 val acc 0.65714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 94  train loss 0.42672 train acc 0.90349 val loss 0.96624 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 95  train loss 0.41935 train acc 0.90413 val loss 0.95582 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 96  train loss 0.41412 train acc 0.90730 val loss 0.94886 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 97  train loss 0.40941 train acc 0.90540 val loss 0.96287 val acc 0.66286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 98  train loss 0.40349 train acc 0.91175 val loss 0.96023 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 99  train loss 0.39739 train acc 0.91048 val loss 0.95370 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 100 train loss 0.39290 train acc 0.91302 val loss 0.96054 val acc 0.66286\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "n_epoch = 100\n",
    "\n",
    "net = ThaiDigitNet() # Initialize the model\n",
    "loss_fn = nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.SGD(net.parameters(), lr=0.01)\n",
    "net, net_history = train(net, n_epoch, loss_fn, optimizer, train_loader, val_loader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "j__w3mymCPJh",
   "metadata": {
    "id": "j__w3mymCPJh"
   },
   "outputs": [],
   "source": [
    "def plot_graph(history):\n",
    "    fig, (ax1, ax2) = plt.subplots(1, 2)\n",
    "    fig.set_figwidth(10)\n",
    "    fig.suptitle(\"Train vs Validation\")\n",
    "    ax1.plot(history[\"train_acc\"], label=\"Train\")\n",
    "    ax1.plot(history[\"val_acc\"], label=\"Validation\")\n",
    "    ax1.legend()\n",
    "    ax1.set_title(\"Accuracy\")\n",
    "\n",
    "    ax2.plot(history[\"train_loss\"], label=\"Train\")\n",
    "    ax2.plot(history[\"val_loss\"], label=\"Validation\")\n",
    "    ax2.legend()\n",
    "    ax2.set_title(\"Loss\")\n",
    "    fig.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "e8dAIHcnCIJA",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 294
    },
    "id": "e8dAIHcnCIJA",
    "outputId": "a0220cd0-b864-4272-f580-77e1b05a5dc2"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 720x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_graph(net_history)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "sMvIE0JbRd0Q",
   "metadata": {
    "id": "sMvIE0JbRd0Q"
   },
   "source": [
    "We can observe \"overfitting\" as the model learns to predict the training set well but perform worse in validation set"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2a0229f1",
   "metadata": {
    "id": "2a0229f1"
   },
   "source": [
    "## **Evaluate the model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "5b41a1c8",
   "metadata": {
    "id": "5b41a1c8"
   },
   "outputs": [],
   "source": [
    "# Create evaluation function for the model\n",
    "def evaluate(val_dir, model):\n",
    "    model.eval()\n",
    "    paths = glob(op.join(val_dir, \"*\", \"*\"))\n",
    "    predictions = []\n",
    "    for idx, path in enumerate(paths):\n",
    "        img = 1 - transform(Image.open(path))\n",
    "        pred = model(img)\n",
    "        predictions.append({\n",
    "            \"path\": path,\n",
    "            \"prediction\": int(pred.argmax(dim=1).item()),\n",
    "            \"label\": int(Path(path).parent.name)\n",
    "        })\n",
    "\n",
    "    results = pd.DataFrame(predictions)\n",
    "    # Calculate accuracy\n",
    "    accuracy = (results.prediction == results.label).sum() / len(results)\n",
    "    print(f\"Accuracy: {accuracy * 100:.2f}%\")\n",
    "    return results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "3WOPs_x_Amap",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 160
    },
    "id": "3WOPs_x_Amap",
    "outputId": "5f5365a2-6277-4910-dcf4-00cea464ffb8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 66.29%\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-21a454f5-bebc-4860-b809-5543bcd48489\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>path</th>\n",
       "      <th>prediction</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>data/validation/8/d86d0ca1-d4cb-48e9-aa8a-6c0f...</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>data/validation/8/23_6db9fc09-96ed-44a9-8052-c...</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>data/validation/8/1e7a0e6c-9b02-434c-bf88-6d60...</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-21a454f5-bebc-4860-b809-5543bcd48489')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-21a454f5-bebc-4860-b809-5543bcd48489 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-21a454f5-bebc-4860-b809-5543bcd48489');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                                                path  prediction  label\n",
       "0  data/validation/8/d86d0ca1-d4cb-48e9-aa8a-6c0f...           8      8\n",
       "1  data/validation/8/23_6db9fc09-96ed-44a9-8052-c...           8      8\n",
       "2  data/validation/8/1e7a0e6c-9b02-434c-bf88-6d60...           8      8"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_df = evaluate(\"data/validation/\", net)\n",
    "pred_df.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "goWDOiKlA0oA",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 281
    },
    "id": "goWDOiKlA0oA",
    "outputId": "83611d86-0a51-499b-88ab-c3043526b788"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Show some random images with their predicted number\n",
    "sample_path = glob(\"data/validation/*/*.png\")[50]\n",
    "img = Image.open(sample_path)\n",
    "img = (1 - transform(img)).squeeze(0)\n",
    "pred = net(img)\n",
    "pred = int(pred.argmax(dim=1))\n",
    "true_class = Path(sample_path).parent.name\n",
    "\n",
    "plt.title(\"Predicted class = {}, True class = {}\".format(pred, true_class))\n",
    "plt.imshow(img, cmap=\"gray\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "-F6H4tFLmUCG",
   "metadata": {
    "id": "-F6H4tFLmUCG"
   },
   "source": [
    "## **Train neural network model with more layers**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "TwMdgtqFI0Lz",
   "metadata": {
    "id": "TwMdgtqFI0Lz"
   },
   "outputs": [],
   "source": [
    "# TODO: Create neural network with more layers\n",
    "class ThaiDigitMoreLayers(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(ThaiDigitMoreLayers, self).__init__()\n",
    "        # TODO: Create layers\n",
    "        self.fc1 = nn.Linear(28 * 28, 392)\n",
    "        self.fc2 = nn.Linear(392, 196)\n",
    "        self.fc3 = nn.Linear(196, 98)\n",
    "        self.fc4 = nn.Linear(98, 10)\n",
    "        \n",
    "    def forward(self, x):\n",
    "        # TODO: forward pass\n",
    "        x = x.view(-1, 28 * 28)\n",
    "        x = self.fc1(x)\n",
    "        x = F.relu(x)\n",
    "        x = self.fc2(x)\n",
    "        x = F.relu(x)\n",
    "        x = self.fc3(x)\n",
    "        x = F.relu(x)\n",
    "        x = self.fc4(x)\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "_dK8OfvaI1P9",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "_dK8OfvaI1P9",
    "outputId": "08551ce1-f5a4-4302-ff52-839295e76b72"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--------------------------------------------------------------------------------\n",
      "Epochs 1   train loss 2.30348 train acc 0.10159 val loss 2.30684 val acc 0.08571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 2   train loss 2.30209 train acc 0.10159 val loss 2.30525 val acc 0.08571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 3   train loss 2.30018 train acc 0.10159 val loss 2.30365 val acc 0.08571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 4   train loss 2.29867 train acc 0.10222 val loss 2.30206 val acc 0.09143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 5   train loss 2.29694 train acc 0.10603 val loss 2.30035 val acc 0.10857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 6   train loss 2.29484 train acc 0.13587 val loss 2.29861 val acc 0.12571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 7   train loss 2.29269 train acc 0.17841 val loss 2.29659 val acc 0.16571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 8   train loss 2.29023 train acc 0.20127 val loss 2.29390 val acc 0.19429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 9   train loss 2.28713 train acc 0.22794 val loss 2.29059 val acc 0.24000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 10  train loss 2.28333 train acc 0.27111 val loss 2.28668 val acc 0.26286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 11  train loss 2.27853 train acc 0.27238 val loss 2.28203 val acc 0.27429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 12  train loss 2.27286 train acc 0.27683 val loss 2.27601 val acc 0.29714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 13  train loss 2.26532 train acc 0.31873 val loss 2.26832 val acc 0.32000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 14  train loss 2.25569 train acc 0.31746 val loss 2.25771 val acc 0.33143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 15  train loss 2.24262 train acc 0.31238 val loss 2.24300 val acc 0.34286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 16  train loss 2.22394 train acc 0.31365 val loss 2.22169 val acc 0.36000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 17  train loss 2.19667 train acc 0.34921 val loss 2.19100 val acc 0.32000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 18  train loss 2.15827 train acc 0.33079 val loss 2.14712 val acc 0.34857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 19  train loss 2.10162 train acc 0.32381 val loss 2.08160 val acc 0.35429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 20  train loss 2.02339 train acc 0.32317 val loss 1.99857 val acc 0.38286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 21  train loss 1.92777 train acc 0.36635 val loss 1.90156 val acc 0.37714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 22  train loss 1.83288 train acc 0.37968 val loss 1.81950 val acc 0.42286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 23  train loss 1.74034 train acc 0.42349 val loss 1.72177 val acc 0.49714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 24  train loss 1.65743 train acc 0.47111 val loss 1.65091 val acc 0.42857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 25  train loss 1.58347 train acc 0.49651 val loss 1.59087 val acc 0.53143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 26  train loss 1.51149 train acc 0.52889 val loss 1.49524 val acc 0.57143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 27  train loss 1.43565 train acc 0.56000 val loss 1.42882 val acc 0.54286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 28  train loss 1.37510 train acc 0.57460 val loss 1.33946 val acc 0.58857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 29  train loss 1.30475 train acc 0.59619 val loss 1.30524 val acc 0.62286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 30  train loss 1.24873 train acc 0.61143 val loss 1.22968 val acc 0.59429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 31  train loss 1.20218 train acc 0.62286 val loss 1.16516 val acc 0.65143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 32  train loss 1.16200 train acc 0.62603 val loss 1.17472 val acc 0.60000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 33  train loss 1.12311 train acc 0.64381 val loss 1.10362 val acc 0.63429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 34  train loss 1.10228 train acc 0.64889 val loss 1.07542 val acc 0.62286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 35  train loss 1.06132 train acc 0.65841 val loss 1.04683 val acc 0.64571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 36  train loss 1.03958 train acc 0.66540 val loss 1.08851 val acc 0.60000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 37  train loss 1.01322 train acc 0.67238 val loss 1.01415 val acc 0.64000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 38  train loss 0.98406 train acc 0.68762 val loss 1.00151 val acc 0.62286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 39  train loss 0.95696 train acc 0.68762 val loss 0.98902 val acc 0.66286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 40  train loss 0.92914 train acc 0.71302 val loss 1.03075 val acc 0.61714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 41  train loss 0.90116 train acc 0.71111 val loss 0.95637 val acc 0.64000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 42  train loss 0.87737 train acc 0.71619 val loss 1.08061 val acc 0.61143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 43  train loss 0.84464 train acc 0.73206 val loss 0.98415 val acc 0.62857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 44  train loss 0.82869 train acc 0.73079 val loss 0.94510 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 45  train loss 0.79712 train acc 0.74984 val loss 0.91567 val acc 0.65143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 46  train loss 0.76589 train acc 0.74921 val loss 0.96066 val acc 0.64000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 47  train loss 0.74207 train acc 0.76444 val loss 0.91743 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 48  train loss 0.71603 train acc 0.76698 val loss 0.91438 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 49  train loss 0.69089 train acc 0.78603 val loss 0.91916 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 50  train loss 0.66661 train acc 0.79175 val loss 0.91181 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 51  train loss 0.64166 train acc 0.79683 val loss 0.88897 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 52  train loss 0.61851 train acc 0.81016 val loss 0.87091 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 53  train loss 0.58633 train acc 0.82286 val loss 0.90017 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 54  train loss 0.56432 train acc 0.82413 val loss 0.87744 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 55  train loss 0.53806 train acc 0.83810 val loss 0.88017 val acc 0.70286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 56  train loss 0.51307 train acc 0.84952 val loss 0.89759 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 57  train loss 0.48815 train acc 0.85460 val loss 0.92533 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 58  train loss 0.46029 train acc 0.86540 val loss 0.93815 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 59  train loss 0.44349 train acc 0.86984 val loss 0.91500 val acc 0.73143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 60  train loss 0.42284 train acc 0.88635 val loss 0.91433 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 61  train loss 0.40265 train acc 0.89333 val loss 0.90216 val acc 0.66286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 62  train loss 0.37047 train acc 0.90667 val loss 0.87254 val acc 0.71429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 63  train loss 0.35259 train acc 0.90667 val loss 0.87254 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 64  train loss 0.32893 train acc 0.92000 val loss 0.91921 val acc 0.70286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 65  train loss 0.31460 train acc 0.92571 val loss 0.96529 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 66  train loss 0.29780 train acc 0.92762 val loss 1.05209 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 67  train loss 0.27643 train acc 0.93714 val loss 0.98533 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 68  train loss 0.26249 train acc 0.94222 val loss 0.92074 val acc 0.74286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 69  train loss 0.24249 train acc 0.94730 val loss 0.94726 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 70  train loss 0.23075 train acc 0.95492 val loss 0.91271 val acc 0.74286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 71  train loss 0.21621 train acc 0.96127 val loss 0.95089 val acc 0.73143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 72  train loss 0.20518 train acc 0.96000 val loss 0.94583 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 73  train loss 0.18754 train acc 0.96952 val loss 1.01853 val acc 0.70857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 74  train loss 0.17949 train acc 0.96825 val loss 0.95828 val acc 0.73143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 75  train loss 0.16494 train acc 0.97524 val loss 0.98030 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 76  train loss 0.15996 train acc 0.97397 val loss 0.96112 val acc 0.74857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 77  train loss 0.14795 train acc 0.97651 val loss 1.01298 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 78  train loss 0.14051 train acc 0.98095 val loss 0.96488 val acc 0.71429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 79  train loss 0.13304 train acc 0.98286 val loss 1.02745 val acc 0.71429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 80  train loss 0.12534 train acc 0.98286 val loss 1.13552 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 81  train loss 0.11491 train acc 0.98476 val loss 1.06384 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 82  train loss 0.10962 train acc 0.98540 val loss 1.03021 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 83  train loss 0.10482 train acc 0.98730 val loss 1.04646 val acc 0.76000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 84  train loss 0.09536 train acc 0.98857 val loss 1.08296 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 85  train loss 0.09206 train acc 0.98921 val loss 1.05725 val acc 0.73143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 86  train loss 0.08524 train acc 0.99048 val loss 1.10399 val acc 0.74286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 87  train loss 0.08134 train acc 0.98984 val loss 1.06862 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 88  train loss 0.07697 train acc 0.98984 val loss 1.11475 val acc 0.73143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 89  train loss 0.07133 train acc 0.99175 val loss 1.14855 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 90  train loss 0.06978 train acc 0.99175 val loss 1.08935 val acc 0.74857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 91  train loss 0.06497 train acc 0.99429 val loss 1.07326 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 92  train loss 0.06154 train acc 0.99492 val loss 1.11350 val acc 0.74286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 93  train loss 0.05867 train acc 0.99556 val loss 1.11054 val acc 0.73143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 94  train loss 0.05610 train acc 0.99619 val loss 1.12772 val acc 0.74857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 95  train loss 0.05252 train acc 0.99683 val loss 1.13437 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 96  train loss 0.05116 train acc 0.99619 val loss 1.17860 val acc 0.73143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 97  train loss 0.04763 train acc 0.99683 val loss 1.23020 val acc 0.70857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 98  train loss 0.04592 train acc 0.99683 val loss 1.19458 val acc 0.71429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 99  train loss 0.04332 train acc 0.99683 val loss 1.19177 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 100 train loss 0.04142 train acc 0.99683 val loss 1.18310 val acc 0.73143\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "n_epochs = 100\n",
    "\n",
    "more_layers_model = ThaiDigitMoreLayers()\n",
    "loss_fn = nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.SGD(more_layers_model.parameters(), lr=0.01)\n",
    "\n",
    "more_layers_model, more_layers_model_history = train(\n",
    "    more_layers_model, n_epochs, loss_fn, optimizer, train_loader, val_loader\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "66MFlYZiHu3U",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 160
    },
    "id": "66MFlYZiHu3U",
    "outputId": "7ddd4794-89a2-4efb-aa4a-6ef74f971528"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 73.14%\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-c88e475c-11ef-44dc-be08-ffb9023925cf\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>path</th>\n",
       "      <th>prediction</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>data/validation/8/d86d0ca1-d4cb-48e9-aa8a-6c0f...</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>data/validation/8/23_6db9fc09-96ed-44a9-8052-c...</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>data/validation/8/1e7a0e6c-9b02-434c-bf88-6d60...</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-c88e475c-11ef-44dc-be08-ffb9023925cf')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-c88e475c-11ef-44dc-be08-ffb9023925cf button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-c88e475c-11ef-44dc-be08-ffb9023925cf');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                                                path  prediction  label\n",
       "0  data/validation/8/d86d0ca1-d4cb-48e9-aa8a-6c0f...           8      8\n",
       "1  data/validation/8/23_6db9fc09-96ed-44a9-8052-c...           8      8\n",
       "2  data/validation/8/1e7a0e6c-9b02-434c-bf88-6d60...           8      8"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_df = evaluate(\"data/validation/\", more_layers_model)\n",
    "pred_df.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "UawDjDZhHwRh",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 294
    },
    "id": "UawDjDZhHwRh",
    "outputId": "a8ec1c0f-d6dc-4978-be9f-4e9d5547acad"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 720x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_graph(more_layers_model_history)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "43ccb6fc",
   "metadata": {
    "id": "43ccb6fc"
   },
   "source": [
    "The training accuracy of the model is higher than the validation accuracy.\n",
    "This means that the model is overfitting the training data.\n",
    "We can try to reduce the overfitting by adding regularization ."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "65fb58d7",
   "metadata": {
    "id": "65fb58d7"
   },
   "source": [
    "## **Regularization with Dropout**\n",
    "\n",
    "Dropout is one of the techniques used to prevent overfitting. Here, we can train the model for more epochs to observe the loss curve."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "cR0MUq3dHXk_",
   "metadata": {
    "id": "cR0MUq3dHXk_"
   },
   "outputs": [],
   "source": [
    "# TODO: Add Dropout layer to the previous neural network\n",
    "class DropoutThaiDigit(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(DropoutThaiDigit, self).__init__()\n",
    "        self.fc1 = nn.Linear(28 * 28, 392)\n",
    "        self.fc2 = nn.Linear(392, 196)\n",
    "        self.fc3 = nn.Linear(196, 98)\n",
    "        self.fc4 = nn.Linear(98, 10)\n",
    "        self.dropout = nn.Dropout(0.1)\n",
    "        \n",
    "    def forward(self, x):\n",
    "        # TODO: forward pass\n",
    "        x = x.view(-1, 28 * 28)\n",
    "        x = self.fc1(x)\n",
    "        x = F.relu(x)\n",
    "        x = self.dropout(x)\n",
    "        x = self.fc2(x)\n",
    "        x = F.relu(x)\n",
    "        x = self.dropout(x)\n",
    "        x = self.fc3(x)\n",
    "        x = F.relu(x)\n",
    "        x = self.dropout(x)\n",
    "        x = self.fc4(x)\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "PEtjdv9TML7F",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "PEtjdv9TML7F",
    "outputId": "2b6d21de-4ad1-4e6a-d729-59d7c051a088"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--------------------------------------------------------------------------------\n",
      "Epochs 1   train loss 2.30427 train acc 0.10222 val loss 2.30502 val acc 0.12000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 2   train loss 2.30317 train acc 0.10032 val loss 2.30416 val acc 0.12000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 3   train loss 2.30139 train acc 0.11556 val loss 2.30305 val acc 0.14857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 4   train loss 2.30017 train acc 0.15429 val loss 2.30209 val acc 0.18286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 5   train loss 2.29853 train acc 0.15937 val loss 2.30089 val acc 0.22857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 6   train loss 2.29704 train acc 0.16508 val loss 2.29940 val acc 0.15429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 7   train loss 2.29533 train acc 0.17270 val loss 2.29778 val acc 0.14857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 8   train loss 2.29332 train acc 0.19175 val loss 2.29590 val acc 0.16571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 9   train loss 2.29051 train acc 0.20444 val loss 2.29360 val acc 0.14286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 10  train loss 2.28823 train acc 0.19175 val loss 2.29053 val acc 0.16571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 11  train loss 2.28425 train acc 0.18730 val loss 2.28649 val acc 0.16571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 12  train loss 2.27941 train acc 0.20254 val loss 2.28111 val acc 0.17714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 13  train loss 2.27399 train acc 0.21397 val loss 2.27441 val acc 0.17714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 14  train loss 2.26548 train acc 0.22222 val loss 2.26482 val acc 0.18857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 15  train loss 2.25341 train acc 0.23810 val loss 2.25148 val acc 0.20571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 16  train loss 2.23801 train acc 0.23619 val loss 2.23281 val acc 0.21714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 17  train loss 2.21542 train acc 0.23873 val loss 2.20610 val acc 0.25143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 18  train loss 2.18169 train acc 0.21524 val loss 2.16719 val acc 0.24000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 19  train loss 2.13792 train acc 0.22032 val loss 2.11508 val acc 0.22857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 20  train loss 2.07653 train acc 0.22667 val loss 2.05251 val acc 0.20000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 21  train loss 2.01512 train acc 0.23048 val loss 1.99032 val acc 0.22286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 22  train loss 1.94617 train acc 0.25778 val loss 1.91865 val acc 0.25143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 23  train loss 1.88204 train acc 0.27238 val loss 1.85549 val acc 0.26286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 24  train loss 1.81505 train acc 0.30984 val loss 1.77322 val acc 0.30857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 25  train loss 1.73977 train acc 0.36762 val loss 1.72513 val acc 0.32571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 26  train loss 1.68605 train acc 0.37778 val loss 1.64527 val acc 0.39429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 27  train loss 1.61928 train acc 0.41270 val loss 1.57380 val acc 0.51429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 28  train loss 1.55522 train acc 0.46349 val loss 1.51934 val acc 0.46857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 29  train loss 1.50328 train acc 0.48762 val loss 1.44619 val acc 0.54857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 30  train loss 1.45258 train acc 0.51429 val loss 1.39544 val acc 0.58286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 31  train loss 1.40154 train acc 0.54413 val loss 1.33586 val acc 0.61714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 32  train loss 1.35036 train acc 0.55048 val loss 1.25865 val acc 0.64571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 33  train loss 1.28849 train acc 0.56571 val loss 1.20921 val acc 0.60571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 34  train loss 1.25722 train acc 0.59048 val loss 1.19294 val acc 0.57714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 35  train loss 1.21638 train acc 0.60635 val loss 1.20704 val acc 0.57143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 36  train loss 1.17226 train acc 0.60889 val loss 1.12191 val acc 0.63429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 37  train loss 1.12839 train acc 0.63556 val loss 1.07542 val acc 0.65143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 38  train loss 1.12109 train acc 0.62603 val loss 1.08625 val acc 0.60571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 39  train loss 1.08785 train acc 0.65270 val loss 1.04035 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 40  train loss 1.05554 train acc 0.65397 val loss 1.03802 val acc 0.64000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 41  train loss 1.02261 train acc 0.66159 val loss 0.99080 val acc 0.64000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 42  train loss 1.00912 train acc 0.66286 val loss 1.01176 val acc 0.64000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 43  train loss 0.98730 train acc 0.67492 val loss 0.96608 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 44  train loss 0.94624 train acc 0.68889 val loss 0.96402 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 45  train loss 0.92863 train acc 0.68952 val loss 0.94762 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 46  train loss 0.91633 train acc 0.69905 val loss 0.95687 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 47  train loss 0.88320 train acc 0.70794 val loss 0.93820 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 48  train loss 0.86387 train acc 0.71810 val loss 0.92695 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 49  train loss 0.86323 train acc 0.71492 val loss 0.92420 val acc 0.67429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 50  train loss 0.81306 train acc 0.74413 val loss 0.91012 val acc 0.70286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 51  train loss 0.77349 train acc 0.74794 val loss 0.91439 val acc 0.71429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 52  train loss 0.76932 train acc 0.74095 val loss 0.90037 val acc 0.70286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 53  train loss 0.74369 train acc 0.76508 val loss 0.88227 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 54  train loss 0.72412 train acc 0.76381 val loss 0.91050 val acc 0.70857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 55  train loss 0.70268 train acc 0.77206 val loss 0.85757 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 56  train loss 0.69540 train acc 0.77397 val loss 0.86015 val acc 0.70857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 57  train loss 0.66620 train acc 0.78540 val loss 0.88431 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 58  train loss 0.63948 train acc 0.80825 val loss 0.90902 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 59  train loss 0.63218 train acc 0.78794 val loss 0.89476 val acc 0.70286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 60  train loss 0.60810 train acc 0.80571 val loss 0.89542 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 61  train loss 0.58036 train acc 0.82159 val loss 0.84335 val acc 0.70857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 62  train loss 0.56937 train acc 0.81905 val loss 0.85152 val acc 0.70286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 63  train loss 0.55487 train acc 0.82730 val loss 0.87001 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 64  train loss 0.52611 train acc 0.82730 val loss 0.86923 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 65  train loss 0.51152 train acc 0.84317 val loss 0.88756 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 66  train loss 0.47016 train acc 0.85524 val loss 0.85866 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 67  train loss 0.47157 train acc 0.85905 val loss 0.88855 val acc 0.74286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 68  train loss 0.45385 train acc 0.86667 val loss 0.86681 val acc 0.70857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 69  train loss 0.43056 train acc 0.87238 val loss 0.85498 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 70  train loss 0.42037 train acc 0.87556 val loss 0.84561 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 71  train loss 0.39417 train acc 0.88698 val loss 0.88382 val acc 0.70857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 72  train loss 0.39841 train acc 0.88889 val loss 0.87170 val acc 0.74286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 73  train loss 0.36641 train acc 0.89079 val loss 0.87595 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 74  train loss 0.35622 train acc 0.89778 val loss 0.87708 val acc 0.76000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 75  train loss 0.33148 train acc 0.91429 val loss 0.87646 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 76  train loss 0.33493 train acc 0.90476 val loss 0.84214 val acc 0.74857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 77  train loss 0.31982 train acc 0.91238 val loss 0.95696 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 78  train loss 0.29260 train acc 0.92698 val loss 0.89725 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 79  train loss 0.27943 train acc 0.93524 val loss 0.87196 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 80  train loss 0.27580 train acc 0.93016 val loss 0.91937 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 81  train loss 0.26984 train acc 0.92635 val loss 0.91600 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 82  train loss 0.26140 train acc 0.92762 val loss 0.97290 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 83  train loss 0.25186 train acc 0.93397 val loss 0.94590 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 84  train loss 0.24211 train acc 0.93651 val loss 1.03383 val acc 0.71429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 85  train loss 0.23441 train acc 0.93841 val loss 0.97694 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 86  train loss 0.22335 train acc 0.94032 val loss 0.92404 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 87  train loss 0.21781 train acc 0.94222 val loss 1.00574 val acc 0.71429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 88  train loss 0.21811 train acc 0.94095 val loss 0.99309 val acc 0.71429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 89  train loss 0.19253 train acc 0.95556 val loss 1.01619 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 90  train loss 0.18915 train acc 0.95365 val loss 0.97720 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 91  train loss 0.18261 train acc 0.95746 val loss 1.03321 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 92  train loss 0.18124 train acc 0.95937 val loss 0.95332 val acc 0.70857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 93  train loss 0.16199 train acc 0.96190 val loss 0.96039 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 94  train loss 0.15725 train acc 0.96698 val loss 0.99355 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 95  train loss 0.15843 train acc 0.96571 val loss 1.00240 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 96  train loss 0.15103 train acc 0.96317 val loss 0.99014 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 97  train loss 0.14205 train acc 0.96444 val loss 0.97297 val acc 0.74286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 98  train loss 0.13088 train acc 0.96952 val loss 1.03878 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 99  train loss 0.13299 train acc 0.97333 val loss 1.01385 val acc 0.70286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 100 train loss 0.13274 train acc 0.97079 val loss 1.00712 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "n_epochs = 100\n",
    "drop_model = DropoutThaiDigit()  # Initialize a model\n",
    "loss_fn = nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.SGD(drop_model.parameters(), lr=0.01)\n",
    "\n",
    "dropout_model, dropout_model_history = train(drop_model, n_epochs, loss_fn, optimizer, train_loader, val_loader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "YrroHtlzO1kn",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 160
    },
    "id": "YrroHtlzO1kn",
    "outputId": "b46e09c1-6179-4e4e-f44a-9a17139ffc2f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 73.71%\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-d1356195-6379-4e33-8983-2b17949f4090\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>path</th>\n",
       "      <th>prediction</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>data/validation/8/d86d0ca1-d4cb-48e9-aa8a-6c0f...</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>data/validation/8/23_6db9fc09-96ed-44a9-8052-c...</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>data/validation/8/1e7a0e6c-9b02-434c-bf88-6d60...</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-d1356195-6379-4e33-8983-2b17949f4090')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-d1356195-6379-4e33-8983-2b17949f4090 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-d1356195-6379-4e33-8983-2b17949f4090');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                                                path  prediction  label\n",
       "0  data/validation/8/d86d0ca1-d4cb-48e9-aa8a-6c0f...           8      8\n",
       "1  data/validation/8/23_6db9fc09-96ed-44a9-8052-c...           8      8\n",
       "2  data/validation/8/1e7a0e6c-9b02-434c-bf88-6d60...           8      8"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_df = evaluate(\"data/validation/\", dropout_model)\n",
    "pred_df.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "1AB10XifO2XD",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 294
    },
    "id": "1AB10XifO2XD",
    "outputId": "ac92290a-699c-4261-ae97-aa5a1252e56d"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 720x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_graph(dropout_model_history)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b611d20b",
   "metadata": {
    "id": "b611d20b"
   },
   "source": [
    "Now the gap between the training accuracy and validation accuracy is smaller than before."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "_R305_c-sBGn",
   "metadata": {
    "id": "_R305_c-sBGn"
   },
   "source": [
    "## **Image Augmentation**\n",
    "\n",
    "Image augmentation is a technique used to artificially increase the size of a training dataset by creating modified versions of existing images."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "rKRgszrQNwPw",
   "metadata": {
    "id": "rKRgszrQNwPw"
   },
   "outputs": [],
   "source": [
    "# TODO: Add image augmentation in `train_transform`\n",
    "\n",
    "train_transform = transforms.Compose([\n",
    "    transforms.Resize((28, 28)),\n",
    "    transforms.Grayscale(),\n",
    "    # add RandomAffine here\n",
    "    transforms.RandomAffine(degrees=(-10, 10), translate=(0.0, 0.1), scale=(1, 1)),\n",
    "    transforms.ToTensor(),\n",
    "])\n",
    "\n",
    "val_transform = transforms.Compose([\n",
    "    transforms.Resize((28, 28)),\n",
    "    transforms.Grayscale(),\n",
    "    transforms.ToTensor(),\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "q6Hq6txiSB_c",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 265
    },
    "id": "q6Hq6txiSB_c",
    "outputId": "ffbe0d2a-4ab8-437c-9237-b81cedbdb465"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPsAAAD4CAYAAAAq5pAIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAPI0lEQVR4nO3df6jVdZ7H8dc7p5umGbpeb9JETVMEsbnNehCjn4s0OELYRIgRk7WxDqEwkxtsuNAISyTbNjFQTekW48SsMpVmhDRTNq27BOGptDR3tzYNFfUqQl6z8sd97x/36+yt7vfzvZ7zPed7uu/nAy73nO/7fM95d7yvvt/z/XzP92PuLgAj3xlVNwCgPQg7EARhB4Ig7EAQhB0I4jvtfDEzSx76nzZtWrtaAZJOnjyZrJ9xRno7aWZltjNsO3fu1MGDB4d88abCbmazJP1K0ihJ/+ruy5p5vnq93szqwLAVDTn39fUl66NHj07Wu7q6TrunMtRqtdxaw7vxZjZK0uOSfiTpckm3mdnljT4fgNZq5jP7dEkfufvH7n5M0mpJc8ppC0DZmgn7+ZJ2Dbq/O1v2FWa2wMzqZsY+OlChlh+gc/flkpZLxQfoALROM1v2PZIuGHT/u9kyAB2ombBvknSpmX3PzLokzZP0UjltAShbw7vx7n7CzBZJ+oMGht6ecfdtpXUGtFBvb2+yfs899yTrc+fOTdbnzZt32j21WlOf2d19vaT1JfUCoIU4XRYIgrADQRB2IAjCDgRB2IEgCDsQRFu/zw6cjv7+/mR9x44dyXpPT09ubcWKFcl1X3/99WT9lltuSdY7EVt2IAjCDgRB2IEgCDsQBGEHgiDsQBAMvbXBkSNHkvVx48Yl60WXNU49/6effppc9/jx48l60VVYjx07lqwfPXo0t7Zv377kum+99Vayvnbt2mT9uuuuy62tXr06uW7R0Nrs2bOT9U7Elh0IgrADQRB2IAjCDgRB2IEgCDsQBGEHgmCcvQRFlyVeunRpsl40Zvvyyy8n61u2bMmtHThwILlu0Th50Tj7iRMnGq4XrVt0/sGMGTOS9TVr1uTWFi5cmFz33nvvTdYnTpyYrHcituxAEIQdCIKwA0EQdiAIwg4EQdiBIAg7EIQVjaOW+mJmyRdrZy+nq6+vL7d23333Jdd9/vnnm3rt7u7uZH3q1Km5tUsuuSS57rnnnpusm1myfvbZZyfr55xzTm6t6L/r4osvTtYvvPDCZP3NN9/MrV177bXJdbu6upL1TlWr1VSv14f8R2vqpBoz2ympT9JJSSfcvdbM8wFonTLOoPsbdz9YwvMAaCE+swNBNBt2l/RHM3vbzBYM9QAzW2BmdTOrN/laAJrQ7G78Ne6+x8wmS3rVzP7L3TcOfoC7L5e0XCo+QAegdZrasrv7nux3r6S1kqaX0RSA8jUcdjMba2bnnLot6YeStpbVGIByNbMb3yNpbTYO+x1J/+bur6RWmDZtmur1aj66F43hP/XUU8n6kiVLcmuLFy9Orrtnz55kff/+/cn6lClTkvVv65hwq82cObPqFjpKw2F3948l/VWJvQBoIYbegCAIOxAEYQeCIOxAEIQdCCLMpaQPHTqUrK9cuTJZnz49/3yhossSjx49Olkv+qomUAa27EAQhB0IgrADQRB2IAjCDgRB2IEgCDsQRJhx9qKvme7YsSNZT33FdcKECQ31BLQTW3YgCMIOBEHYgSAIOxAEYQeCIOxAEIQdCCLMOPvhw4eT9ePHjyfrRZdzRvmK/k2OHj2arBdNRx0NW3YgCMIOBEHYgSAIOxAEYQeCIOxAEIQdCCLMOPuXX36ZrPf39yfrY8aMKbMdSDpy5Eiy/uCDDybrb7zxRrK+cePG3NqZZ56ZXHckKtyym9kzZtZrZlsHLZtoZq+a2YfZb67eAHS44ezG/0bSrK8tu1/SBne/VNKG7D6ADlYYdnffKOnrcyfNkXRqvqSVkm4uuS8AJWv0AF2Pu+/Nbu+T1JP3QDNbYGZ1M6sfOHCgwZcD0Kymj8a7u0vyRH25u9fcvdbd3d3sywFoUKNh329mUyQp+91bXksAWqHRsL8kaX52e76kdeW0A6BVCsfZzWyVpBskTTKz3ZJ+IWmZpN+b2d2SPpE0t5VNlqFoXNXMkvWicXqcvt7e9A7hqlWrkvWiaww88cQTubVFixYl1x01alSy/m1UGHZ3vy2nNLPkXgC0EKfLAkEQdiAIwg4EQdiBIAg7EESYr7iOHz8+WS8amhupp/oOnACZ7+TJk8l60RBVakhz8uTJyXVvuummZP2xxx5L1k+cOJFbu+OOO5LrjsRpuNmyA0EQdiAIwg4EQdiBIAg7EARhB4Ig7EAQYcbZJ02alKxPnDgxWd+yZUturdmx6FYq+mruc889l6yvX78+Wb/++uuT9dR49rhx45LrPvTQQ8n6s88+m6x//vnnubVjx44l1x2J2LIDQRB2IAjCDgRB2IEgCDsQBGEHgiDsQBBhxtl7enJnqJIkXXXVVcn6a6+9llvbtm1bct2pU6cm681KjfMXjaMvXrw4WR87dmyy/sorryTrqesAFF3OuegcgdmzZyfrqX+XM86It52L918MBEXYgSAIOxAEYQeCIOxAEIQdCIKwA0GEGWcvui78nXfemaxv2LAht3b77bcn13333XeT9SJ9fX3J+osvvphbe+CBB5LrTps2LVlftmxZsv74448n64888khuLfWeSlJ/f39T9fPOOy+3VnT+wEhUuGU3s2fMrNfMtg5attTM9pjZ5uwnfXYDgMoNZzf+N5JmDbH8UXe/MvtJX84EQOUKw+7uGyUdakMvAFqomQN0i8zsvWw3P3diLDNbYGZ1M6uP1PnSgG+DRsP+a0nfl3SlpL2Sco/CuPtyd6+5e627u7vBlwPQrIbC7u773f2ku/dLWiFperltAShbQ2E3symD7v5Y0ta8xwLoDIXj7Ga2StINkiaZ2W5Jv5B0g5ldKckl7ZT00xb22BZXX311sv7kk0/m1p5++unkukVzgReNF+/atStZ3759e25txowZyXUfffTRZP2yyy5L1h9++OFk/YorrsitrVu3LrluV1dXsr5p06ZkPXWOwJgxY5LrjkSFYXf324ZYnP7rBtBxOF0WCIKwA0EQdiAIwg4EQdiBIMzd2/ZitVrN6/V6216vTKnLGn/xxRfJdWfNGup7RP8v9VVMSbrxxhuT9VtvvTW3Nnny5OS6rZb6+/rss8+S6xZd7rloyPOuu+7KrRVNF/1tVavVVK/XbagaW3YgCMIOBEHYgSAIOxAEYQeCIOxAEIQdCIJx9jY4fPhwsn7WWWcl60Vf9TQbclgVI1TRv7e7M84OREbYgSAIOxAEYQeCIOxAEIQdCIKwA0GEmbK5SuPHj6+6BYAtOxAFYQeCIOxAEIQdCIKwA0EQdiAIwg4EQdiBIArDbmYXmNmfzOwDM9tmZj/Llk80s1fN7MPs94TWtwugUcPZsp+Q9PfufrmkGZIWmtnlku6XtMHdL5W0IbsPoEMVht3d97r7O9ntPknbJZ0vaY6kldnDVkq6uVVNAmjeaX1mN7OLJP1A0luSetx9b1baJ6knZ50FZlY3s/qBAweaaBVAM4YddjMbJ+kFST93969cQdEHrlo55JUr3X25u9fcvdbd3d1UswAaN6ywm9mZGgj679x9TbZ4v5lNyepTJPW2pkUAZRjO0XiT9LSk7e7+y0GllyTNz27Pl7Su/PYAlGU432e/WtJPJL1vZpuzZUskLZP0ezO7W9Inkua2pkUAZSgMu7v/p6S8q9LPLLcdAK3CGXRAEIQdCIKwA0EQdiAIwg4EQdiBIAg7EARhB4Ig7EAQhB0IgrADQRB2IAjCDgRB2IEgCDsQBGEHgiDsQBCEHQiCsANBEHYgCMIOBEHYgSAIOxAEYQeCIOxAEIQdCIKwA0EQdiAIwg4EYe6efoDZBZJ+K6lHkkta7u6/MrOlkv5O0oHsoUvcfX3Bc6VfbIQqeo+B02GWN6nyAHcf8gHDCfsUSVPc/R0zO0fS25Ju1sB87Efc/V9Oo8mQf/WEHWVqNOzDmZ99r6S92e0+M9su6fwGegRQodP6zG5mF0n6gaS3skWLzOw9M3vGzCbkrLPAzOpmVm+qUwBNKdyN//MDzcZJ+ndJD7r7GjPrkXRQA5/j/0kDu/p/W/AcIfdn2Y1HmVr2mT178jMlvSzpD+7+yyHqF0l62d3/suB5Qv7VE3aUqdGwF+7G28AzPy1p++CgZwfuTvmxpK3D6hRAJYZzNP4aSf8h6X1J/dniJZJuk3SlBnbjd0r6aXYwL/VcbOJQmqh7TC3djS8LYUeZCPvQGt6NBzAyEHYgCMIOBEHYgSAIOxAEYQeCKPwiDNCpioag8FVs2YEgCDsQBGEHgiDsQBCEHQiCsANBEHYgiHaPsx+U9Mmg+5OyZZ2oU3vr1L4kemtUmb1dmFdo6/fZv/HiZnV3r1XWQEKn9tapfUn01qh29cZuPBAEYQeCqDrsyyt+/ZRO7a1T+5LorVFt6a3Sz+wA2qfqLTuANiHsQBCVhN3MZpnZf5vZR2Z2fxU95DGznWb2vpltrnp+umwOvV4z2zpo2UQze9XMPsx+DznHXkW9LTWzPdl7t9nMZlfU2wVm9icz+8DMtpnZz7Lllb53ib7a8r61/TO7mY2S9D+SbpS0W9ImSbe5+wdtbSSHme2UVHP3yk/AMLPrJB2R9NtTU2uZ2T9LOuTuy7L/UU5w93/okN6W6jSn8W5Rb3nTjN+pCt+7Mqc/b0QVW/bpkj5y94/d/Zik1ZLmVNBHx3P3jZIOfW3xHEkrs9srNfDH0nY5vXUEd9/r7u9kt/sknZpmvNL3LtFXW1QR9vMl7Rp0f7c6a753l/RHM3vbzBZU3cwQegZNs7VPUk+VzQyhcBrvdvraNOMd8941Mv15szhA903XuPtfS/qRpIXZ7mpH8oHPYJ00dvprSd/XwByAeyU9UmUz2TTjL0j6ubsfHlyr8r0boq+2vG9VhH2PpAsG3f9utqwjuPue7HevpLUa+NjRSfafmkE3+91bcT9/5u773f2ku/dLWqEK37tsmvEXJP3O3ddkiyt/74bqq13vWxVh3yTpUjP7npl1SZon6aUK+vgGMxubHTiRmY2V9EN13lTUL0man92eL2ldhb18RadM4503zbgqfu8qn/7c3dv+I2m2Bo7I/6+kf6yih5y+Lpa0JfvZVnVvklZpYLfuuAaObdwt6S8kbZD0oaTXJE3soN6e1cDU3u9pIFhTKurtGg3sor8naXP2M7vq9y7RV1veN06XBYLgAB0QBGEHgiDsQBCEHQiCsANBEHYgCMIOBPF/7je+f/w9DlQAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "path = glob(\"deep-learning-skooldio/thai-handwritten-dataset/*/*\")[30]\n",
    "img = Image.open(path)\n",
    "\n",
    "# TODO: Transform image using `train_transform`\n",
    "transformed_img = train_transform(img).squeeze(0)\n",
    "\n",
    "plt.imshow(transformed_img, cmap=\"gray\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "7UrwfkQVTer2",
   "metadata": {
    "id": "7UrwfkQVTer2"
   },
   "outputs": [],
   "source": [
    "train_thaidigit_dataset = ThaiDigitDataset(\"data/train/\", transform=train_transform)\n",
    "val_thaidigit_dataset = ThaiDigitDataset(\"data/validation/\", transform=val_transform)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "Ovkv1UXwTdd1",
   "metadata": {
    "id": "Ovkv1UXwTdd1"
   },
   "outputs": [],
   "source": [
    "train_dataloader = DataLoader(train_thaidigit_dataset, batch_size=16, shuffle=True)\n",
    "val_dataloader = DataLoader(val_thaidigit_dataset, batch_size=16, shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "d0OHv65iJgHs",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "d0OHv65iJgHs",
    "outputId": "6165fc66-516e-4299-c267-ae251ded005b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--------------------------------------------------------------------------------\n",
      "Epochs 1   train loss 2.30356 train acc 0.10667 val loss 2.30306 val acc 0.10857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 2   train loss 2.30207 train acc 0.10984 val loss 2.30252 val acc 0.10286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 3   train loss 2.30086 train acc 0.11492 val loss 2.30200 val acc 0.13143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 4   train loss 2.30005 train acc 0.13206 val loss 2.30142 val acc 0.10857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 5   train loss 2.29875 train acc 0.11746 val loss 2.30062 val acc 0.12000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 6   train loss 2.29730 train acc 0.12063 val loss 2.29975 val acc 0.13143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 7   train loss 2.29636 train acc 0.13079 val loss 2.29871 val acc 0.13143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 8   train loss 2.29536 train acc 0.13079 val loss 2.29724 val acc 0.15429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 9   train loss 2.29429 train acc 0.14603 val loss 2.29584 val acc 0.14857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 10  train loss 2.29170 train acc 0.15492 val loss 2.29414 val acc 0.14286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 11  train loss 2.29039 train acc 0.16825 val loss 2.29229 val acc 0.14286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 12  train loss 2.28845 train acc 0.16000 val loss 2.28989 val acc 0.14857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 13  train loss 2.28381 train acc 0.15556 val loss 2.28694 val acc 0.19429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 14  train loss 2.28108 train acc 0.16571 val loss 2.28294 val acc 0.25143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 15  train loss 2.27578 train acc 0.20698 val loss 2.27812 val acc 0.27429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 16  train loss 2.27049 train acc 0.21841 val loss 2.27206 val acc 0.27429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 17  train loss 2.26130 train acc 0.22413 val loss 2.26274 val acc 0.30286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 18  train loss 2.25189 train acc 0.21841 val loss 2.25098 val acc 0.35429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 19  train loss 2.23831 train acc 0.25587 val loss 2.23581 val acc 0.35429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 20  train loss 2.21998 train acc 0.27175 val loss 2.21480 val acc 0.36000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 21  train loss 2.19613 train acc 0.29143 val loss 2.18552 val acc 0.36571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 22  train loss 2.16085 train acc 0.29460 val loss 2.14422 val acc 0.36571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 23  train loss 2.10895 train acc 0.29206 val loss 2.08560 val acc 0.38286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 24  train loss 2.04457 train acc 0.32127 val loss 2.01098 val acc 0.44000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 25  train loss 1.94967 train acc 0.33587 val loss 1.90755 val acc 0.43429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 26  train loss 1.86369 train acc 0.36571 val loss 1.81447 val acc 0.43429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 27  train loss 1.77751 train acc 0.39429 val loss 1.72485 val acc 0.46286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 28  train loss 1.70712 train acc 0.39937 val loss 1.63386 val acc 0.49143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 29  train loss 1.62518 train acc 0.43746 val loss 1.59957 val acc 0.47429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 30  train loss 1.59615 train acc 0.44571 val loss 1.51462 val acc 0.52000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 31  train loss 1.56578 train acc 0.45968 val loss 1.49684 val acc 0.48000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 32  train loss 1.53371 train acc 0.45968 val loss 1.45267 val acc 0.50857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 33  train loss 1.49610 train acc 0.47429 val loss 1.42781 val acc 0.55429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 34  train loss 1.44973 train acc 0.49841 val loss 1.40142 val acc 0.53143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 35  train loss 1.43724 train acc 0.51810 val loss 1.38075 val acc 0.56571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 36  train loss 1.42244 train acc 0.51048 val loss 1.36114 val acc 0.57143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 37  train loss 1.41738 train acc 0.50921 val loss 1.36686 val acc 0.57714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 38  train loss 1.38949 train acc 0.53079 val loss 1.32314 val acc 0.59429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 39  train loss 1.36511 train acc 0.55111 val loss 1.28393 val acc 0.58286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 40  train loss 1.35438 train acc 0.54032 val loss 1.23574 val acc 0.64000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 41  train loss 1.31639 train acc 0.57460 val loss 1.25221 val acc 0.59429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 42  train loss 1.27991 train acc 0.57778 val loss 1.24403 val acc 0.61714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 43  train loss 1.27213 train acc 0.58095 val loss 1.18211 val acc 0.60000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 44  train loss 1.28608 train acc 0.57397 val loss 1.20788 val acc 0.61143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 45  train loss 1.24276 train acc 0.58603 val loss 1.20551 val acc 0.63429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 46  train loss 1.21035 train acc 0.60444 val loss 1.16037 val acc 0.66286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 47  train loss 1.20464 train acc 0.59873 val loss 1.17043 val acc 0.65714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 48  train loss 1.18844 train acc 0.60698 val loss 1.16117 val acc 0.64000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 49  train loss 1.17516 train acc 0.61270 val loss 1.13282 val acc 0.64571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 50  train loss 1.13783 train acc 0.62984 val loss 1.15941 val acc 0.61714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 51  train loss 1.13735 train acc 0.63873 val loss 1.10610 val acc 0.64000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 52  train loss 1.12347 train acc 0.63111 val loss 1.10786 val acc 0.66286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 53  train loss 1.09213 train acc 0.64635 val loss 1.13587 val acc 0.65143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 54  train loss 1.09035 train acc 0.63810 val loss 1.05629 val acc 0.64000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 55  train loss 1.06976 train acc 0.65397 val loss 1.05889 val acc 0.65143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 56  train loss 1.05838 train acc 0.66032 val loss 1.02784 val acc 0.66286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 57  train loss 1.02522 train acc 0.64635 val loss 1.03205 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 58  train loss 1.02734 train acc 0.66413 val loss 1.02595 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 59  train loss 0.99947 train acc 0.66667 val loss 1.07002 val acc 0.68000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 60  train loss 1.02022 train acc 0.67175 val loss 0.97918 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 61  train loss 1.02057 train acc 0.66730 val loss 1.05724 val acc 0.65143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 62  train loss 0.95810 train acc 0.67492 val loss 0.99229 val acc 0.68571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 63  train loss 0.96462 train acc 0.67937 val loss 0.98905 val acc 0.66857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 64  train loss 0.95939 train acc 0.67175 val loss 1.02238 val acc 0.62857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 65  train loss 0.97049 train acc 0.67683 val loss 0.95025 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 66  train loss 0.92461 train acc 0.70349 val loss 0.92328 val acc 0.71429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 67  train loss 0.93943 train acc 0.68635 val loss 0.95101 val acc 0.70286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 68  train loss 0.93300 train acc 0.69397 val loss 0.87932 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 69  train loss 0.91265 train acc 0.70286 val loss 0.91963 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 70  train loss 0.91547 train acc 0.69778 val loss 0.96667 val acc 0.65143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 71  train loss 0.89707 train acc 0.71048 val loss 0.87993 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 72  train loss 0.85429 train acc 0.72063 val loss 0.88489 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 73  train loss 0.88533 train acc 0.71111 val loss 0.95773 val acc 0.69143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 74  train loss 0.86667 train acc 0.70857 val loss 0.91863 val acc 0.66286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 75  train loss 0.86770 train acc 0.71746 val loss 0.88671 val acc 0.70286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 76  train loss 0.85455 train acc 0.72381 val loss 0.86595 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 77  train loss 0.83036 train acc 0.73143 val loss 0.86964 val acc 0.70857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 78  train loss 0.80922 train acc 0.73016 val loss 0.85046 val acc 0.69714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 79  train loss 0.82981 train acc 0.72127 val loss 0.86664 val acc 0.71429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 80  train loss 0.82497 train acc 0.71683 val loss 0.81899 val acc 0.73143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 81  train loss 0.83283 train acc 0.72381 val loss 0.82407 val acc 0.70857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 82  train loss 0.78092 train acc 0.73651 val loss 0.79245 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 83  train loss 0.78754 train acc 0.73714 val loss 0.80770 val acc 0.71429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 84  train loss 0.76876 train acc 0.73841 val loss 0.76764 val acc 0.70857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 85  train loss 0.79066 train acc 0.74286 val loss 0.80338 val acc 0.70286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 86  train loss 0.77461 train acc 0.74286 val loss 0.74013 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 87  train loss 0.76798 train acc 0.74984 val loss 0.75365 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 88  train loss 0.75478 train acc 0.74286 val loss 0.75105 val acc 0.73143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 89  train loss 0.73086 train acc 0.77206 val loss 0.76855 val acc 0.72000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 90  train loss 0.71789 train acc 0.75683 val loss 0.77214 val acc 0.70857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 91  train loss 0.71617 train acc 0.76571 val loss 0.76474 val acc 0.74286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 92  train loss 0.72567 train acc 0.75492 val loss 0.76167 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 93  train loss 0.70737 train acc 0.78349 val loss 0.73515 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 94  train loss 0.70821 train acc 0.76381 val loss 0.79513 val acc 0.73143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 95  train loss 0.70254 train acc 0.76952 val loss 0.67552 val acc 0.74857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 96  train loss 0.70308 train acc 0.76508 val loss 0.76383 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 97  train loss 0.68371 train acc 0.78095 val loss 0.71006 val acc 0.72571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 98  train loss 0.67956 train acc 0.78476 val loss 0.69529 val acc 0.76571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 99  train loss 0.65697 train acc 0.78476 val loss 0.71058 val acc 0.75429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 100 train loss 0.64643 train acc 0.78159 val loss 0.71835 val acc 0.73143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 101 train loss 0.65782 train acc 0.78095 val loss 0.70292 val acc 0.74857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 102 train loss 0.66624 train acc 0.78032 val loss 0.62256 val acc 0.77714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 103 train loss 0.64353 train acc 0.78095 val loss 0.66935 val acc 0.74286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 104 train loss 0.62818 train acc 0.79111 val loss 0.69381 val acc 0.76000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 105 train loss 0.62528 train acc 0.79746 val loss 0.69214 val acc 0.74286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 106 train loss 0.63267 train acc 0.79492 val loss 0.63458 val acc 0.77714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 107 train loss 0.63830 train acc 0.79048 val loss 0.70394 val acc 0.76000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 108 train loss 0.62137 train acc 0.78730 val loss 0.69361 val acc 0.75429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 109 train loss 0.61051 train acc 0.80127 val loss 0.68106 val acc 0.73714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 110 train loss 0.61372 train acc 0.80000 val loss 0.68317 val acc 0.75429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 111 train loss 0.60969 train acc 0.79365 val loss 0.68870 val acc 0.76000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 112 train loss 0.60505 train acc 0.80000 val loss 0.66427 val acc 0.75429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 113 train loss 0.58295 train acc 0.80000 val loss 0.64133 val acc 0.76571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 114 train loss 0.55784 train acc 0.80635 val loss 0.62182 val acc 0.77714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 115 train loss 0.55427 train acc 0.82095 val loss 0.69148 val acc 0.74857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 116 train loss 0.57687 train acc 0.80508 val loss 0.63780 val acc 0.76571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 117 train loss 0.54348 train acc 0.82603 val loss 0.67305 val acc 0.76000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 118 train loss 0.56515 train acc 0.81079 val loss 0.65165 val acc 0.77143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 119 train loss 0.55948 train acc 0.81206 val loss 0.62711 val acc 0.78286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 120 train loss 0.56050 train acc 0.80762 val loss 0.65379 val acc 0.77143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 121 train loss 0.56413 train acc 0.80825 val loss 0.61455 val acc 0.78286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 122 train loss 0.55261 train acc 0.80571 val loss 0.59331 val acc 0.78857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 123 train loss 0.54507 train acc 0.82222 val loss 0.59655 val acc 0.78286\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 124 train loss 0.52811 train acc 0.82730 val loss 0.61750 val acc 0.80000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 125 train loss 0.52562 train acc 0.82286 val loss 0.60825 val acc 0.78857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 126 train loss 0.51997 train acc 0.82476 val loss 0.63601 val acc 0.77714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 127 train loss 0.53097 train acc 0.81714 val loss 0.60025 val acc 0.80000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 128 train loss 0.50769 train acc 0.82984 val loss 0.57236 val acc 0.81143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 129 train loss 0.51247 train acc 0.83683 val loss 0.58022 val acc 0.79429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 130 train loss 0.52206 train acc 0.81968 val loss 0.61118 val acc 0.78857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 131 train loss 0.50858 train acc 0.83556 val loss 0.57686 val acc 0.80000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 132 train loss 0.52439 train acc 0.82984 val loss 0.60339 val acc 0.81143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 133 train loss 0.48969 train acc 0.84444 val loss 0.58956 val acc 0.81143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 134 train loss 0.49043 train acc 0.84190 val loss 0.56541 val acc 0.80000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 135 train loss 0.49054 train acc 0.83810 val loss 0.55690 val acc 0.81143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 136 train loss 0.49312 train acc 0.83746 val loss 0.54913 val acc 0.79429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 137 train loss 0.51214 train acc 0.83302 val loss 0.58071 val acc 0.81143\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 138 train loss 0.45691 train acc 0.84698 val loss 0.61097 val acc 0.80000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 139 train loss 0.47781 train acc 0.83746 val loss 0.59166 val acc 0.80000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 140 train loss 0.47672 train acc 0.85524 val loss 0.56825 val acc 0.81714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 141 train loss 0.48726 train acc 0.83429 val loss 0.63964 val acc 0.76571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 142 train loss 0.46188 train acc 0.85460 val loss 0.60984 val acc 0.79429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 143 train loss 0.46359 train acc 0.84825 val loss 0.58012 val acc 0.80000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 144 train loss 0.44533 train acc 0.85841 val loss 0.61951 val acc 0.78857\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 145 train loss 0.46552 train acc 0.85270 val loss 0.55957 val acc 0.81714\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 146 train loss 0.44261 train acc 0.84571 val loss 0.58311 val acc 0.80571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 147 train loss 0.47588 train acc 0.83810 val loss 0.54795 val acc 0.79429\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 148 train loss 0.43554 train acc 0.86222 val loss 0.56987 val acc 0.80000\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 149 train loss 0.46606 train acc 0.84698 val loss 0.59073 val acc 0.80571\n",
      "--------------------------------------------------------------------------------\n",
      "Epochs 150 train loss 0.44114 train acc 0.85714 val loss 0.57094 val acc 0.78857\n",
      "--------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "# Let's train the model with augmented data\n",
    "n_epochs = 150\n",
    "augmented_model = DropoutThaiDigit()  # Initialize the model\n",
    "loss_fn = nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.SGD(augmented_model.parameters(), lr=0.01)\n",
    "\n",
    "augmented_model, augmented_model_history = train(\n",
    "    augmented_model, n_epochs, loss_fn, optimizer, train_dataloader, val_dataloader\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "0_lbTJOXJheG",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 160
    },
    "id": "0_lbTJOXJheG",
    "outputId": "fbe5a54f-1891-48e6-cced-1dd9bcf5d1cd"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 78.86%\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-56de0128-e2af-4e43-8fd0-6825179ff2d1\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>path</th>\n",
       "      <th>prediction</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>data/validation/8/d86d0ca1-d4cb-48e9-aa8a-6c0f...</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>data/validation/8/23_6db9fc09-96ed-44a9-8052-c...</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>data/validation/8/1e7a0e6c-9b02-434c-bf88-6d60...</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-56de0128-e2af-4e43-8fd0-6825179ff2d1')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-56de0128-e2af-4e43-8fd0-6825179ff2d1 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-56de0128-e2af-4e43-8fd0-6825179ff2d1');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                                                path  prediction  label\n",
       "0  data/validation/8/d86d0ca1-d4cb-48e9-aa8a-6c0f...           8      8\n",
       "1  data/validation/8/23_6db9fc09-96ed-44a9-8052-c...           8      8\n",
       "2  data/validation/8/1e7a0e6c-9b02-434c-bf88-6d60...           8      8"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Evaluate the new model with old dataset.\n",
    "pred_df = evaluate(\"data/validation/\", augmented_model)\n",
    "pred_df.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "1kByhfphJi4M",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 294
    },
    "id": "1kByhfphJi4M",
    "outputId": "3953e0f2-3cd8-4ed7-cc51-8487eb4db618"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 720x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_graph(augmented_model_history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "GPwNfAB_Jj7L",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 279
    },
    "id": "GPwNfAB_Jj7L",
    "outputId": "bf92dd54-7a9c-481b-916a-8dee84850121"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Confusion matrix\n",
    "cm = confusion_matrix(pred_df.label, pred_df.prediction, labels=range(10))\n",
    "display = ConfusionMatrixDisplay(confusion_matrix=cm, display_labels=range(10))\n",
    "display.plot()\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "provenance": []
  },
  "gpuClass": "standard",
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.6"
  },
  "vscode": {
   "interpreter": {
    "hash": "31f2aee4e71d21fbe5cf8b01ff0e069b9275f58929596ceb00d14d90e3e16cd6"
   }
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "04bff9d1c6b44aba879b1c1bd99eacd3": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_e585b2c5a15649c1ba70dab769b0baa0",
       "IPY_MODEL_95666913c37848a7995ceff186e516a1",
       "IPY_MODEL_35ad23534198412db6e4233a01d34a61"
      ],
      "layout": "IPY_MODEL_986f13b440ef4477b4a026ddca6e42c0"
     }
    },
    "1b225e76df10484181b6b41a6a587d00": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "35ad23534198412db6e4233a01d34a61": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_a950a090716e4dfca165a5ac2eb898d3",
      "placeholder": "​",
      "style": "IPY_MODEL_73b451e0cc17454db355df0008c5b78e",
      "value": " 1575/1575 [00:00&lt;00:00, 7055.18it/s]"
     }
    },
    "45ab66db880c4c02a326ef3bf7c7c43c": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "546987ef32044cf0a967ee6cf5ff567b": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "561963521ebb40cdad9789b56981f431": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_a0e3edff6bbc4871b46e8aa80479973a",
      "placeholder": "​",
      "style": "IPY_MODEL_872049e09056436692c61b449c1b2769",
      "value": " 175/175 [00:00&lt;00:00, 3220.92it/s]"
     }
    },
    "73b451e0cc17454db355df0008c5b78e": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "755bcd54d24e4787a6875b0a004c3536": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "792e5f6e426b4dcc98914f0502c207c4": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_1b225e76df10484181b6b41a6a587d00",
      "max": 175,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_755bcd54d24e4787a6875b0a004c3536",
      "value": 175
     }
    },
    "83df746b0eee4a9eba9783a903db697c": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "872049e09056436692c61b449c1b2769": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "8aaa4f0f5ac54220b650540a0a6c4d6f": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "8d69cd19ce0141e1bb6cf76c94b3fd0f": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_83df746b0eee4a9eba9783a903db697c",
      "placeholder": "​",
      "style": "IPY_MODEL_99b59fba156b4c1d99bb8d0c23eeba16",
      "value": "100%"
     }
    },
    "95666913c37848a7995ceff186e516a1": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_c1005f423da84e93ba9fdb1788ecb9da",
      "max": 1575,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_45ab66db880c4c02a326ef3bf7c7c43c",
      "value": 1575
     }
    },
    "986f13b440ef4477b4a026ddca6e42c0": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "99b59fba156b4c1d99bb8d0c23eeba16": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "a0e3edff6bbc4871b46e8aa80479973a": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "a950a090716e4dfca165a5ac2eb898d3": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "c1005f423da84e93ba9fdb1788ecb9da": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "e33ce989c88d4e1da8aedc4a988e55fb": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "e585b2c5a15649c1ba70dab769b0baa0": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_546987ef32044cf0a967ee6cf5ff567b",
      "placeholder": "​",
      "style": "IPY_MODEL_e33ce989c88d4e1da8aedc4a988e55fb",
      "value": "100%"
     }
    },
    "eb9c6b8de45c47dc8fac7cad28d2f6d9": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_8d69cd19ce0141e1bb6cf76c94b3fd0f",
       "IPY_MODEL_792e5f6e426b4dcc98914f0502c207c4",
       "IPY_MODEL_561963521ebb40cdad9789b56981f431"
      ],
      "layout": "IPY_MODEL_8aaa4f0f5ac54220b650540a0a6c4d6f"
     }
    }
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}