01_train.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Google Colabで動かす場合\n",
    "!pip install extra_keras_datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 🚩 Step:1-1\n",
    "import numpy as np\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import tensorflow as tf\n",
    "import tensorflow.keras as keras\n",
    "import extra_keras_datasets.emnist as emnist\n",
    "\n",
    "from tensorflow.keras.layers import Flatten, Dense\n",
    "from tensorflow.keras.layers import Conv2D, MaxPooling2D\n",
    "from tensorflow.keras.layers import BatchNormalization, Dropout\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "numpy               : 1.19.0\n",
      "matplotlib          : 3.3.0\n",
      "tensorflow          : 2.2.0\n",
      "tensorflow.keras    : 2.3.0-tf\n"
     ]
    }
   ],
   "source": [
    "# 主要パッケージのバージョン\n",
    "def show_version(pkg):\n",
    "    print(\"{0:20}: {1}\".format(pkg.__name__, pkg.__version__))\n",
    "\n",
    "\n",
    "# 🚩 Step:1-2\n",
    "show_version(np)\n",
    "show_version(matplotlib)\n",
    "show_version(tf)\n",
    "show_version(keras)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 🚩 Step:1-3\n",
    "# EMNIST-Lettersのデータをロード(初回はダウンロードに数分かかる)\n",
    "(t_images, t_labels), (v_images, v_labels) = emnist.load_data(type=\"letters\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "t_images    : uint8, (124800, 28, 28)\n",
      "t_labels    : uint8, (124800,)\n",
      "v_images    : uint8, (20800, 28, 28)\n",
      "v_labels    : uint8, (20800,)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 取得データの形状を確認\n",
    "def show_data_info(label, data):\n",
    "    print(\"{:12}: {}, {}\".format(label, data.dtype, data.shape))\n",
    "\n",
    "\n",
    "# Matplotlib 3.3.0より前のimshow互換用 (h,w,1)→(h,w)\n",
    "def compatible_image(img):\n",
    "    if img.ndim == 3 and img.shape[2] == 1:\n",
    "        img = img.reshape(img.shape[0], img.shape[1])\n",
    "    return img\n",
    "\n",
    "\n",
    "# 取得データ画像のサンプル表示\n",
    "def show_image(img):\n",
    "    plt.rcParams[\"font.size\"] = 10\n",
    "    plt.figure()\n",
    "    plt.imshow(compatible_image(img))\n",
    "    plt.colorbar()\n",
    "    plt.grid(False)\n",
    "    plt.show()\n",
    "\n",
    "\n",
    "# 🚩 Step:1-4\n",
    "show_data_info(\"t_images\", t_images)\n",
    "show_data_info(\"t_labels\", t_labels)\n",
    "show_data_info(\"v_images\", v_images)\n",
    "show_data_info(\"v_labels\", v_labels)\n",
    "show_image(t_images[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# データ形式を4次元データ (batch, height, width, channel)に変換\n",
    "# 各ピクセルの情報も0..255から0..1の範囲にノーマライズ\n",
    "def adjust_image(img):\n",
    "    img = img.astype(\"float32\")\n",
    "    img = img / 255.0\n",
    "    img = img.reshape(-1, 28, 28, 1)  # 3D→4D\n",
    "    return img\n",
    "\n",
    "\n",
    "# 🚩 Step:1-5\n",
    "t_images = adjust_image(t_images)\n",
    "v_images = adjust_image(v_images)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "t_images    : float32, (124800, 28, 28, 1)\n",
      "v_images    : float32, (20800, 28, 28, 1)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "show_data_info(\"t_images\", t_images)\n",
    "show_data_info(\"v_images\", v_images)\n",
    "show_image(t_images[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 28, 28, 32)        320       \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 28, 28, 32)        9248      \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 28, 28, 32)        25632     \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 14, 14, 32)        0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 14, 14, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 14, 14, 64)        18496     \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 14, 14, 64)        36928     \n",
      "_________________________________________________________________\n",
      "conv2d_5 (Conv2D)            (None, 14, 14, 64)        102464    \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 7, 7, 64)          0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 7, 7, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 3136)              0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 128)               401536    \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 27)                3483      \n",
      "=================================================================\n",
      "Total params: 598,107\n",
      "Trainable params: 598,107\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# 🚩 Step:1-6\n",
    "# モデル構築\n",
    "# 畳み込み、プーリング、ドロップアウトの組み合わせで、最後は完全結合でsoftmaxでクラス分け\n",
    "# 隠れ層の活性化関数には定番のReLUを使用\n",
    "# TensorFlowを使う場合、最初の隠れ層だけ入力データの形式をinput_shapeで指定する必要がある\n",
    "model = keras.Sequential(\n",
    "    [\n",
    "        Conv2D(32, (3, 3), activation=\"relu\", padding=\"same\", input_shape=(28, 28, 1)),\n",
    "        Conv2D(32, (3, 3), activation=\"relu\", padding=\"same\"),\n",
    "        Conv2D(32, (5, 5), activation=\"relu\", padding=\"same\"),\n",
    "        MaxPooling2D((2, 2), padding=\"same\"),\n",
    "        Dropout(0.4),\n",
    "        Conv2D(64, (3, 3), activation=\"relu\", padding=\"same\"),\n",
    "        Conv2D(64, (3, 3), activation=\"relu\", padding=\"same\"),\n",
    "        Conv2D(64, (5, 5), activation=\"relu\", padding=\"same\"),\n",
    "        MaxPooling2D((2, 2), padding=\"same\"),\n",
    "        Dropout(0.4),\n",
    "        Flatten(),\n",
    "        Dense(128, activation=\"relu\"),\n",
    "        Dropout(0.4),\n",
    "        Dense(27, activation=\"softmax\"),\n",
    "    ]\n",
    ")\n",
    "\n",
    "# 🚩 Step:1-7\n",
    "# モデルのコンパイル\n",
    "opt = keras.optimizers.Adam()\n",
    "loss = keras.losses.SparseCategoricalCrossentropy()\n",
    "model.compile(optimizer=opt, loss=loss, metrics=[\"accuracy\"])\n",
    "\n",
    "# モデルの概要\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "975/975 - 6s - loss: 0.6501 - accuracy: 0.8004 - val_loss: 0.2181 - val_accuracy: 0.9260\n",
      "Epoch 2/20\n",
      "975/975 - 6s - loss: 0.2898 - accuracy: 0.9077 - val_loss: 0.1846 - val_accuracy: 0.9390\n",
      "Epoch 3/20\n",
      "975/975 - 6s - loss: 0.2463 - accuracy: 0.9201 - val_loss: 0.1717 - val_accuracy: 0.9410\n",
      "Epoch 4/20\n",
      "975/975 - 6s - loss: 0.2204 - accuracy: 0.9280 - val_loss: 0.1684 - val_accuracy: 0.9435\n",
      "Epoch 5/20\n",
      "975/975 - 6s - loss: 0.2057 - accuracy: 0.9324 - val_loss: 0.1612 - val_accuracy: 0.9460\n",
      "Epoch 6/20\n",
      "975/975 - 6s - loss: 0.1931 - accuracy: 0.9357 - val_loss: 0.1556 - val_accuracy: 0.9481\n",
      "Epoch 7/20\n",
      "975/975 - 6s - loss: 0.1849 - accuracy: 0.9372 - val_loss: 0.1573 - val_accuracy: 0.9476\n",
      "Epoch 8/20\n",
      "975/975 - 6s - loss: 0.1781 - accuracy: 0.9393 - val_loss: 0.1547 - val_accuracy: 0.9482\n",
      "Epoch 9/20\n",
      "975/975 - 6s - loss: 0.1706 - accuracy: 0.9419 - val_loss: 0.1568 - val_accuracy: 0.9486\n",
      "Epoch 10/20\n",
      "975/975 - 6s - loss: 0.1670 - accuracy: 0.9425 - val_loss: 0.1488 - val_accuracy: 0.9508\n",
      "Epoch 11/20\n",
      "975/975 - 6s - loss: 0.1626 - accuracy: 0.9447 - val_loss: 0.1556 - val_accuracy: 0.9466\n",
      "Epoch 12/20\n",
      "975/975 - 6s - loss: 0.1565 - accuracy: 0.9448 - val_loss: 0.1559 - val_accuracy: 0.9488\n",
      "Epoch 13/20\n",
      "975/975 - 6s - loss: 0.1541 - accuracy: 0.9455 - val_loss: 0.1514 - val_accuracy: 0.9500\n",
      "Epoch 14/20\n",
      "975/975 - 6s - loss: 0.1496 - accuracy: 0.9476 - val_loss: 0.1539 - val_accuracy: 0.9506\n",
      "Epoch 15/20\n",
      "975/975 - 6s - loss: 0.1470 - accuracy: 0.9483 - val_loss: 0.1536 - val_accuracy: 0.9492\n",
      "Epoch 16/20\n",
      "975/975 - 6s - loss: 0.1452 - accuracy: 0.9482 - val_loss: 0.1475 - val_accuracy: 0.9508\n",
      "Epoch 17/20\n",
      "975/975 - 6s - loss: 0.1419 - accuracy: 0.9493 - val_loss: 0.1498 - val_accuracy: 0.9498\n",
      "Epoch 18/20\n",
      "975/975 - 6s - loss: 0.1417 - accuracy: 0.9497 - val_loss: 0.1477 - val_accuracy: 0.9508\n",
      "Epoch 19/20\n",
      "975/975 - 6s - loss: 0.1388 - accuracy: 0.9505 - val_loss: 0.1482 - val_accuracy: 0.9505\n",
      "Epoch 20/20\n",
      "975/975 - 6s - loss: 0.1357 - accuracy: 0.9503 - val_loss: 0.1547 - val_accuracy: 0.9516\n"
     ]
    }
   ],
   "source": [
    "# 🚩 Step:1-8\n",
    "# 教育データで訓練（GPUが利用できない場合は計算に時間がかかる）\n",
    "history = model.fit(\n",
    "    t_images,\n",
    "    t_labels,\n",
    "    epochs=20,\n",
    "    batch_size=128,\n",
    "    validation_data=(v_images, v_labels),\n",
    "    verbose=2,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# エポックごとの学習結果の推移を確認\n",
    "# TensorBoardを使えば、更に色々な情報を確認できる\n",
    "def show_history(history):\n",
    "    plt.rcParams[\"font.size\"] = 10\n",
    "    fig, ax = plt.subplots(1, 2, figsize=(10, 5))\n",
    "    t_col = \"#03F\"\n",
    "    v_col = \"#F33\"\n",
    "\n",
    "    ax[0].set_title(\"Accuracy\", fontsize=14)\n",
    "    ax[0].plot(history.history[\"accuracy\"], color=t_col, label=\"accuracy\")\n",
    "    ax[0].plot(history.history[\"val_accuracy\"], color=v_col, label=\"val_accuracy\")\n",
    "    ax[1].set_title(\"Loss\", fontsize=14)\n",
    "    ax[1].plot(history.history[\"loss\"], color=t_col, label=\"loss\")\n",
    "    ax[1].plot(history.history[\"val_loss\"], color=v_col, label=\"val_loss\")\n",
    "\n",
    "    x_pos = np.arange(0, len(history.epoch), 2)\n",
    "    for x in ax:\n",
    "        x.set_xticks(x_pos)\n",
    "        x.set_xticklabels(x_pos + 1)\n",
    "        x.grid(True)\n",
    "        x.legend(fontsize=12)\n",
    "\n",
    "    plt.show()\n",
    "\n",
    "\n",
    "# 🚩 Step:1-9\n",
    "show_history(history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 🚩 Step:1-10\n",
    "# 学習済みモデルの保存（HDF5形式）\n",
    "model.save(\"data/emnist-model.h5\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}