TrainFacialExpression.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "from keras.applications.vgg16 import VGG16\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.models import Sequential, Model\n",
    "from keras.layers import Input, Activation, Dropout, Flatten, Dense\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras import optimizers\n",
    "import numpy as np\n",
    "\n",
    "os.environ['KMP_DUPLICATE_LIB_OK']='TRUE'\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"]=\"0\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 分類するクラス\n",
    "classes = ['angry', 'disgust', 'fear', 'happy', 'neutral', 'sad', 'surprise']\n",
    "\n",
    "nb_classes = len(classes)\n",
    "\n",
    "img_width, img_height = 48, 48\n",
    "\n",
    "# トレーニング用とバリデーション用の画像格納先\n",
    "train_data_dir = './data2/train'\n",
    "validation_data_dir = './data2/validation'\n",
    "\n",
    "\n",
    "nb_train_samples = 2100\n",
    "nb_validation_samples = 210\n",
    "\n",
    "batch_size = 32\n",
    "nb_epoch = 50"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def vgg_model_maker():\n",
    "    \"\"\" VGG16のモデルをFC層以外使用。FC層のみ作成して結合して用意する \"\"\"\n",
    "\n",
    "    # VGG16のロード。FC層は不要なので include_top=False\n",
    "    input_tensor = Input(shape=(img_width, img_height, 3))\n",
    "    vgg16 = VGG16(include_top=False, weights='imagenet', input_tensor=input_tensor)\n",
    "\n",
    "    # FC層の作成\n",
    "    top_model = Sequential()\n",
    "    top_model.add(Flatten(input_shape=vgg16.output_shape[1:]))\n",
    "    top_model.add(Dense(256, activation='relu'))\n",
    "    top_model.add(Dropout(0.5))\n",
    "    top_model.add(Dense(nb_classes, activation='softmax'))\n",
    "\n",
    "    # VGG16とFC層を結合してモデルを作成\n",
    "    model = Model(inputs=vgg16.input, outputs=top_model(vgg16.output))\n",
    "\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def image_generator():\n",
    "    \"\"\" ディレクトリ内の画像を読み込んでトレーニングデータとバリデーションデータの作成 \"\"\"\n",
    "    train_datagen = ImageDataGenerator(\n",
    "        rescale=1.0 / 255,\n",
    "        zoom_range=0.2,\n",
    "        horizontal_flip=True)\n",
    "\n",
    "    validation_datagen = ImageDataGenerator(rescale=1.0 / 255)\n",
    "\n",
    "    train_generator = train_datagen.flow_from_directory(\n",
    "        train_data_dir,\n",
    "        target_size=(img_width, img_height),\n",
    "        color_mode='rgb',\n",
    "        classes=classes,\n",
    "        class_mode='categorical',\n",
    "        batch_size=batch_size,\n",
    "        shuffle=True)\n",
    "\n",
    "    validation_generator = validation_datagen.flow_from_directory(\n",
    "        validation_data_dir,\n",
    "        target_size=(img_width, img_height),\n",
    "        color_mode='rgb',\n",
    "        classes=classes,\n",
    "        class_mode='categorical',\n",
    "        batch_size=batch_size,\n",
    "        shuffle=True)\n",
    "\n",
    "    return (train_generator, validation_generator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "result_dir = 'trainingResult'\n",
    "if not os.path.exists(result_dir):\n",
    "    os.mkdir(result_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 2100 images belonging to 7 classes.\n",
      "Found 210 images belonging to 7 classes.\n",
      "Epoch 1/50\n",
      "65/65 [==============================] - 104s 2s/step - loss: 2.0048 - accuracy: 0.1683 - val_loss: 1.8724 - val_accuracy: 0.2135\n",
      "Epoch 2/50\n",
      "65/65 [==============================] - 104s 2s/step - loss: 1.9076 - accuracy: 0.2041 - val_loss: 1.9290 - val_accuracy: 0.2472\n",
      "Epoch 3/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.8570 - accuracy: 0.2268 - val_loss: 1.8708 - val_accuracy: 0.2697\n",
      "Epoch 4/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.8225 - accuracy: 0.2495 - val_loss: 1.8125 - val_accuracy: 0.2865\n",
      "Epoch 5/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.8005 - accuracy: 0.2708 - val_loss: 1.7958 - val_accuracy: 0.2865\n",
      "Epoch 6/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.7737 - accuracy: 0.2935 - val_loss: 1.8052 - val_accuracy: 0.2809\n",
      "Epoch 7/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.7495 - accuracy: 0.2993 - val_loss: 1.7614 - val_accuracy: 0.3090\n",
      "Epoch 8/50\n",
      "65/65 [==============================] - 104s 2s/step - loss: 1.7025 - accuracy: 0.3346 - val_loss: 1.6640 - val_accuracy: 0.3490\n",
      "Epoch 9/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.6923 - accuracy: 0.3366 - val_loss: 1.7055 - val_accuracy: 0.3596\n",
      "Epoch 10/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.6326 - accuracy: 0.3615 - val_loss: 1.3658 - val_accuracy: 0.3989\n",
      "Epoch 11/50\n",
      "65/65 [==============================] - 134s 2s/step - loss: 1.6455 - accuracy: 0.3487 - val_loss: 1.5452 - val_accuracy: 0.4213\n",
      "Epoch 12/50\n",
      "65/65 [==============================] - 104s 2s/step - loss: 1.5902 - accuracy: 0.3774 - val_loss: 1.4918 - val_accuracy: 0.4831\n",
      "Epoch 13/50\n",
      "65/65 [==============================] - 102s 2s/step - loss: 1.5938 - accuracy: 0.3862 - val_loss: 1.6708 - val_accuracy: 0.3539\n",
      "Epoch 14/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.5687 - accuracy: 0.3951 - val_loss: 1.8346 - val_accuracy: 0.5056\n",
      "Epoch 15/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.5219 - accuracy: 0.4221 - val_loss: 1.4718 - val_accuracy: 0.4792\n",
      "Epoch 16/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.4914 - accuracy: 0.4367 - val_loss: 1.3834 - val_accuracy: 0.4944\n",
      "Epoch 17/50\n",
      "65/65 [==============================] - 104s 2s/step - loss: 1.4605 - accuracy: 0.4500 - val_loss: 1.3126 - val_accuracy: 0.4831\n",
      "Epoch 18/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.4319 - accuracy: 0.4480 - val_loss: 1.3423 - val_accuracy: 0.4944\n",
      "Epoch 19/50\n",
      "65/65 [==============================] - 104s 2s/step - loss: 1.3522 - accuracy: 0.4885 - val_loss: 1.3225 - val_accuracy: 0.5337\n",
      "Epoch 20/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.3683 - accuracy: 0.4835 - val_loss: 1.1798 - val_accuracy: 0.5674\n",
      "Epoch 21/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.3433 - accuracy: 0.5024 - val_loss: 1.3099 - val_accuracy: 0.5730\n",
      "Epoch 22/50\n",
      "65/65 [==============================] - 104s 2s/step - loss: 1.3092 - accuracy: 0.5102 - val_loss: 1.3190 - val_accuracy: 0.5208\n",
      "Epoch 23/50\n",
      "65/65 [==============================] - 1523s 23s/step - loss: 1.2591 - accuracy: 0.5343 - val_loss: 1.0172 - val_accuracy: 0.6124\n",
      "Epoch 24/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.2412 - accuracy: 0.5343 - val_loss: 1.0238 - val_accuracy: 0.5056\n",
      "Epoch 25/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.1864 - accuracy: 0.5672 - val_loss: 1.1139 - val_accuracy: 0.6966\n",
      "Epoch 26/50\n",
      "65/65 [==============================] - 104s 2s/step - loss: 1.1778 - accuracy: 0.5644 - val_loss: 0.9307 - val_accuracy: 0.6124\n",
      "Epoch 27/50\n",
      "65/65 [==============================] - 102s 2s/step - loss: 1.1310 - accuracy: 0.5880 - val_loss: 0.9328 - val_accuracy: 0.6348\n",
      "Epoch 28/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 1.0258 - accuracy: 0.6248 - val_loss: 0.9599 - val_accuracy: 0.7191\n",
      "Epoch 29/50\n",
      "65/65 [==============================] - 106s 2s/step - loss: 1.0546 - accuracy: 0.6308 - val_loss: 0.5928 - val_accuracy: 0.7344\n",
      "Epoch 30/50\n",
      "65/65 [==============================] - 105s 2s/step - loss: 0.9996 - accuracy: 0.6347 - val_loss: 0.7292 - val_accuracy: 0.7809\n",
      "Epoch 31/50\n",
      "65/65 [==============================] - 103s 2s/step - loss: 0.9420 - accuracy: 0.6610 - val_loss: 1.0125 - val_accuracy: 0.7022\n",
      "Epoch 32/50\n",
      "65/65 [==============================] - 814s 13s/step - loss: 0.9433 - accuracy: 0.6524 - val_loss: 0.6221 - val_accuracy: 0.7921\n",
      "Epoch 33/50\n",
      "65/65 [==============================] - 107s 2s/step - loss: 0.8863 - accuracy: 0.6862 - val_loss: 0.5307 - val_accuracy: 0.7921\n",
      "Epoch 34/50\n",
      "65/65 [==============================] - 106s 2s/step - loss: 0.8533 - accuracy: 0.6863 - val_loss: 0.6669 - val_accuracy: 0.8371\n",
      "Epoch 35/50\n",
      "65/65 [==============================] - 110s 2s/step - loss: 0.8262 - accuracy: 0.7074 - val_loss: 0.5849 - val_accuracy: 0.7584\n",
      "Epoch 36/50\n",
      "65/65 [==============================] - 108s 2s/step - loss: 0.7804 - accuracy: 0.7263 - val_loss: 0.8547 - val_accuracy: 0.7292\n",
      "Epoch 37/50\n",
      "65/65 [==============================] - 106s 2s/step - loss: 0.7420 - accuracy: 0.7447 - val_loss: 0.5859 - val_accuracy: 0.8708\n",
      "Epoch 38/50\n",
      "65/65 [==============================] - 113s 2s/step - loss: 0.7189 - accuracy: 0.7495 - val_loss: 0.5477 - val_accuracy: 0.8764\n",
      "Epoch 39/50\n",
      "65/65 [==============================] - 108s 2s/step - loss: 0.6921 - accuracy: 0.7500 - val_loss: 0.2378 - val_accuracy: 0.8820\n",
      "Epoch 40/50\n",
      "65/65 [==============================] - 110s 2s/step - loss: 0.6712 - accuracy: 0.7596 - val_loss: 0.3939 - val_accuracy: 0.9157\n",
      "Epoch 41/50\n",
      "65/65 [==============================] - 113s 2s/step - loss: 0.7151 - accuracy: 0.7393 - val_loss: 0.6062 - val_accuracy: 0.8539\n",
      "Epoch 42/50\n",
      "65/65 [==============================] - 106s 2s/step - loss: 0.5635 - accuracy: 0.8058 - val_loss: 0.2920 - val_accuracy: 0.9213\n",
      "Epoch 43/50\n",
      "65/65 [==============================] - 114s 2s/step - loss: 0.5931 - accuracy: 0.7882 - val_loss: 0.2966 - val_accuracy: 0.8802\n",
      "Epoch 44/50\n",
      "65/65 [==============================] - 112s 2s/step - loss: 0.5842 - accuracy: 0.8022 - val_loss: 0.2942 - val_accuracy: 0.8764\n",
      "Epoch 45/50\n",
      "65/65 [==============================] - 107s 2s/step - loss: 0.4661 - accuracy: 0.8428 - val_loss: 0.2807 - val_accuracy: 0.9213\n",
      "Epoch 46/50\n",
      "65/65 [==============================] - 115s 2s/step - loss: 0.4579 - accuracy: 0.8380 - val_loss: 0.3383 - val_accuracy: 0.9045\n",
      "Epoch 47/50\n",
      "65/65 [==============================] - 111s 2s/step - loss: 0.4439 - accuracy: 0.8496 - val_loss: 0.4053 - val_accuracy: 0.9270\n",
      "Epoch 48/50\n",
      "65/65 [==============================] - 111s 2s/step - loss: 0.3958 - accuracy: 0.8643 - val_loss: 0.2756 - val_accuracy: 0.9382\n",
      "Epoch 49/50\n",
      "65/65 [==============================] - 116s 2s/step - loss: 0.4140 - accuracy: 0.8567 - val_loss: 0.2430 - val_accuracy: 0.9270\n",
      "Epoch 50/50\n",
      "65/65 [==============================] - 107s 2s/step - loss: 0.4089 - accuracy: 0.8677 - val_loss: 0.1853 - val_accuracy: 0.9375\n"
     ]
    }
   ],
   "source": [
    "# モデル作成\n",
    "vgg_model = vgg_model_maker()\n",
    "\n",
    "# 最後のconv層の直前までの層をfreeze\n",
    "for layer in vgg_model.layers[:15]:\n",
    "    layer.trainable = False\n",
    "\n",
    "# 多クラス分類を指定\n",
    "vgg_model.compile(loss='categorical_crossentropy',\n",
    "                  optimizer=optimizers.SGD(lr=1e-3, momentum=0.9),\n",
    "                metrics=['accuracy'])\n",
    "\n",
    "# 画像のジェネレータ生成\n",
    "train_generator, validation_generator = image_generator()\n",
    "\n",
    "# Fine-tuning\n",
    "\n",
    "history = vgg_model.fit_generator(\n",
    "    train_generator,\n",
    "    steps_per_epoch=int(nb_train_samples/batch_size),\n",
    "    epochs=nb_epoch,\n",
    "    validation_data=validation_generator,\n",
    "    validation_steps=int(nb_validation_samples/batch_size)\n",
    ")\n",
    "\n",
    "vgg_model.save_weights(os.path.join(result_dir, 'finetuning-60-2.h5'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "loss     = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "nb_epoch = len(loss)\n",
    "plt.plot(range(nb_epoch), loss,     marker='.', label='loss')\n",
    "plt.plot(range(nb_epoch), val_loss, marker='.', label='val_loss')\n",
    "plt.legend(loc='best', fontsize=10)\n",
    "plt.grid()\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss     = history.history['accuracy']\n",
    "val_loss = history.history['val_accuracy']\n",
    "\n",
    "nb_epoch = len(loss)\n",
    "plt.plot(range(nb_epoch), loss,     marker='.', label='accuracy')\n",
    "plt.plot(range(nb_epoch), val_loss, marker='.', label='val_accuracy')\n",
    "plt.legend(loc='best', fontsize=10)\n",
    "plt.grid()\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('accuracy')\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}