01-02.train_model.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "name": "01-02.train_model.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Y7Oh08yAhHwT"
      },
      "source": [
        "# XXXXXXXX"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_OwRCliygZBI",
        "outputId": "dca5cb19-8beb-4a7c-fa10-73d9d4765665"
      },
      "source": [
        "!git clone https://github.com/Kazuhito00/7-segment-display-reader"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Cloning into '7-segment-display-reader'...\n",
            "remote: Enumerating objects: 21857, done.\u001b[K\n",
            "remote: Counting objects: 100% (21857/21857), done.\u001b[K\n",
            "remote: Compressing objects: 100% (21844/21844), done.\u001b[K\n",
            "remote: Total 21857 (delta 13), reused 21852 (delta 12), pack-reused 0\u001b[K\n",
            "Receiving objects: 100% (21857/21857), 58.41 MiB | 24.56 MiB/s, done.\n",
            "Resolving deltas: 100% (13/13), done.\n",
            "Checking out files: 100% (41994/41994), done.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "B6iLw5KChHAk"
      },
      "source": [
        "# XXXXXXXX"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8hJhtoVbtXtc",
        "outputId": "f493e9c0-0537-4a1b-af9d-f1c0268f282e"
      },
      "source": [
        "!git clone https://github.com/Kazuhito00/7seg-image-generator.git"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Cloning into '7seg-image-generator'...\n",
            "remote: Enumerating objects: 45, done.\u001b[K\n",
            "remote: Counting objects: 100% (45/45), done.\u001b[K\n",
            "remote: Compressing objects: 100% (36/36), done.\u001b[K\n",
            "remote: Total 45 (delta 21), reused 20 (delta 6), pack-reused 0\u001b[K\n",
            "Unpacking objects: 100% (45/45), done.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UjwC9CxN2fwZ",
        "outputId": "14339cb8-b9ee-4ff2-de7d-12f5090f1c48"
      },
      "source": [
        "!python '7seg-image-generator/create_7segment_dataset_da(easy).py' \\\n",
        "    --erase_debug_window \\\n",
        "    --steps=4000 \\\n",
        "    --start_count=10000000"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100% 4000/4000 [00:44<00:00, 90.65it/s]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "A3-dkAwMia4v"
      },
      "source": [
        "# XXXXXXXX"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "k0rX5nAe9hTO"
      },
      "source": [
        "%cp -rf './7-segment-display-reader/01.dataset/00' './dataset'\n",
        "%cp -rf './7-segment-display-reader/01.dataset/01' './dataset'\n",
        "%cp -rf './7-segment-display-reader/01.dataset/02' './dataset'\n",
        "%cp -rf './7-segment-display-reader/01.dataset/03' './dataset'\n",
        "%cp -rf './7-segment-display-reader/01.dataset/04' './dataset'\n",
        "%cp -rf './7-segment-display-reader/01.dataset/05' './dataset'\n",
        "%cp -rf './7-segment-display-reader/01.dataset/06' './dataset'\n",
        "%cp -rf './7-segment-display-reader/01.dataset/07' './dataset'\n",
        "%cp -rf './7-segment-display-reader/01.dataset/08' './dataset'\n",
        "%cp -rf './7-segment-display-reader/01.dataset/09' './dataset'\n",
        "%cp -rf './7-segment-display-reader/01.dataset/11' './dataset'"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SJTAw24d3WvH"
      },
      "source": [
        "# XXXXXXXX"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 286
        },
        "id": "5BQUP7Vx1GtC",
        "outputId": "eb834869-cef9-40e8-9852-5dfa20dc560c"
      },
      "source": [
        "import cv2\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "test_image = cv2.imread('./dataset/00/00000000.jpg')\n",
        "test_image = cv2.resize(test_image, (96, 96))\n",
        "test_image = cv2.cvtColor(test_image, cv2.COLOR_BGR2RGB)\n",
        " \n",
        "plt.imshow(test_image)"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f443c024290>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 5
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Jha2pIhn3Kum",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 286
        },
        "outputId": "abf5a374-d4b5-442e-e4d2-04adf32e8f55"
      },
      "source": [
        "import cv2\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "test_image = cv2.imread('./dataset/00/00000000.jpg')\n",
        "test_image = cv2.resize(test_image, (96, 96))\n",
        "\n",
        "# グレースケール化\n",
        "test_image = cv2.cvtColor(test_image, cv2.COLOR_BGR2GRAY)\n",
        "\n",
        "# # ヒストグラム平坦化\n",
        "clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))\n",
        "test_image = clahe.apply(test_image)\n",
        "\n",
        "# # ガウシアンブラー\n",
        "test_image = cv2.GaussianBlur(test_image, (3, 3), 0)\n",
        "\n",
        "# 2値化\n",
        "_, test_image = cv2.threshold(test_image, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)\n",
        "\n",
        "test_image = cv2.cvtColor(test_image, cv2.COLOR_GRAY2RGB)\n",
        " \n",
        "plt.imshow(test_image)"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f442c6c9150>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jXu-uksG5sp9"
      },
      "source": [
        "import os\n",
        "\n",
        "dataset_directory = './dataset'\n",
        "dataset2_directory = './dataset2'\n",
        "\n",
        "# 変換後データ格納ディレクトリ作成　※「dataset_directory」と同様の構成\n",
        "for class_dir in os.listdir(dataset_directory):\n",
        "    os.makedirs(dataset2_directory + '/' + class_dir, exist_ok=True)"
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "STm7F3oU6ZjV",
        "outputId": "353020f7-6b75-460a-80af-fc49d738e1d7"
      },
      "source": [
        "from tqdm import tqdm\n",
        "\n",
        "for class_dir in tqdm(os.listdir(dataset_directory)):\n",
        "    for filename in os.listdir(dataset_directory + '/' + class_dir):\n",
        "        file_path = os.path.join(dataset_directory, class_dir, filename)\n",
        "        save_path = os.path.join(dataset2_directory, class_dir, filename)\n",
        "        \n",
        "        test_image = cv2.imread(file_path)\n",
        "        test_image = cv2.resize(test_image, (96, 96))\n",
        "\n",
        "        # グレースケール化\n",
        "        test_image = cv2.cvtColor(test_image, cv2.COLOR_BGR2GRAY)\n",
        "\n",
        "        # ヒストグラム平坦化\n",
        "        clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))\n",
        "        test_image = clahe.apply(test_image)\n",
        "\n",
        "        # ガウシアンブラー\n",
        "        test_image = cv2.GaussianBlur(test_image, (3, 3), 0)\n",
        "\n",
        "        # 2値化\n",
        "        _, test_image = cv2.threshold(test_image, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)\n",
        "\n",
        "        # BGR形式へ戻す\n",
        "        test_image = cv2.cvtColor(test_image, cv2.COLOR_GRAY2BGR)\n",
        "        \n",
        "        # 保存\n",
        "        cv2.imwrite(save_path, test_image)"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 12/12 [01:00<00:00,  5.03s/it]\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BUuk1DIvifE7"
      },
      "source": [
        "# XXXXXXXX"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tJo-iE4twmkr"
      },
      "source": [
        "import os\n",
        "\n",
        "dataset_directory = './dataset2'\n",
        "train_directory = './train'\n",
        "validation_directory = './validation'\n",
        "\n",
        "# 学習データ格納ディレクトリ作成　※「dataset_directory」と同様の構成\n",
        "for dir_path in os.listdir(dataset_directory):\n",
        "    os.makedirs(train_directory + '/' + dir_path, exist_ok=True)\n",
        "# 検証データ格納ディレクトリ作成　※「dataset_directory」と同様の構成\n",
        "for dir_path in os.listdir(dataset_directory):\n",
        "    os.makedirs(validation_directory + '/' + dir_path, exist_ok=True)"
      ],
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CRiXbfC2hRg7"
      },
      "source": [
        "import os\n",
        "import random\n",
        "\n",
        "import numpy as np\n",
        "import tensorflow as tf\n",
        "\n",
        "seed = 42\n",
        "\n",
        "random.seed(seed)\n",
        "np.random.seed(seed)\n",
        "tf.random.set_seed(seed)\n",
        "os.environ[\"PYTHONHASHSEED\"] = str(seed)"
      ],
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "plq6tEfKv6KC"
      },
      "source": [
        "import glob\n",
        "import shutil\n",
        "import random\n",
        "\n",
        "train_ratio = 0.75  # 学習データの割合\n",
        "\n",
        "random.seed(42)\n",
        "\n",
        "# コピー元ディレクトリ取得\n",
        "directory_list = glob.glob(dataset_directory + '/*')\n",
        "for temp_directory in directory_list:\n",
        "    file_list = glob.glob(temp_directory + '/*')\n",
        "\n",
        "    # ディレクトリへコピー\n",
        "    for index, filepath in enumerate(file_list):\n",
        "        if index < int(len(file_list) * train_ratio):\n",
        "            # 学習用データ\n",
        "            shutil.copy2(filepath, train_directory + '/' + os.path.basename(temp_directory))\n",
        "        else:\n",
        "            # 検証用データ\n",
        "            shutil.copy2(filepath, validation_directory + '/' + os.path.basename(temp_directory))"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "v4ugse-pvgxS"
      },
      "source": [
        "# XXXXXXX"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VAaCp9vO21FS"
      },
      "source": [
        "!pip install -U albumentations"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8i5VFmLy20WF"
      },
      "source": [
        "# Albumentationsを用いたデータ拡張設定\n",
        "import albumentations as A\n",
        "\n",
        "def preprocessing_augmentation_function(param_p = 0.0):\n",
        "    transform = [\n",
        "        A.Cutout(num_holes=8,\n",
        "                 max_h_size=8,\n",
        "                 max_w_size=8,\n",
        "                 fill_value=0,\n",
        "                 p=param_p),\n",
        "        A.Cutout(num_holes=8,\n",
        "                 max_h_size=8,\n",
        "                 max_w_size=8,\n",
        "                 fill_value=255,\n",
        "                 p=param_p),\n",
        "    ]\n",
        "    augmentation_function = A.Compose(transform)\n",
        "\n",
        "    def augmentation(x):\n",
        "        augmentation_image = augmentation_function(image=x)\n",
        "        return augmentation_image['image']\n",
        "\n",
        "    return augmentation"
      ],
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pfpACsOZ-5MH"
      },
      "source": [
        "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
        "\n",
        "train_image_da_generator = ImageDataGenerator(\n",
        "    rescale=1.0/255,\n",
        "    preprocessing_function=preprocessing_augmentation_function(0.1),\n",
        ")\n",
        "validation_image_generator = ImageDataGenerator(rescale=1.0/255)"
      ],
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PwigwiYUaWBO",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9d569ca2-982d-4b9f-c797-7a3e4c005032"
      },
      "source": [
        "batch_size = 64\n",
        "image_height, image_width = 96, 96\n",
        "\n",
        "train_data_gen = train_image_da_generator.flow_from_directory(\n",
        "    batch_size=batch_size,\n",
        "    directory=train_directory,\n",
        "    shuffle=True,\n",
        "    target_size=(image_height, image_width),\n",
        "    class_mode='categorical'\n",
        ")\n",
        "validation_data_gen = validation_image_generator.flow_from_directory(\n",
        "    batch_size=batch_size,\n",
        "    directory=validation_directory,\n",
        "    shuffle=False,\n",
        "    target_size=(image_height, image_width),\n",
        "    class_mode='categorical'\n",
        ")"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 67490 images belonging to 12 classes.\n",
            "Found 22500 images belonging to 12 classes.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "v7Flgo8v3Y0w",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "3b91b11e-6053-4df6-b342-87c992de9b31"
      },
      "source": [
        "base_model = tf.keras.applications.MobileNetV2(include_top=False, weights='imagenet', input_shape=(96, 96, 3))\n",
        "\n",
        "base_model.trainable = True\n",
        "\n",
        "x = tf.keras.layers.GlobalAveragePooling2D()(base_model.output)\n",
        "output = tf.keras.layers.Dense(12, activation='softmax', name='last_output')(x)\n",
        "model = tf.keras.Model(inputs=base_model.inputs, outputs=output, name='model')\n",
        "\n",
        "model.compile(\n",
        "    optimizer='sgd',\n",
        "    loss='categorical_crossentropy',\n",
        "    metrics=['accuracy']\n",
        ")"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/mobilenet_v2/mobilenet_v2_weights_tf_dim_ordering_tf_kernels_1.0_96_no_top.h5\n",
            "9412608/9406464 [==============================] - 0s 0us/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MIrx4g4V4Zdi"
      },
      "source": [
        "# モデルチェックポイント保存用コールバック\n",
        "checkpoint_path = os.path.join(os.getcwd(), 'checkpoints', 'weights.hdf5')\n",
        "cp_callback = tf.keras.callbacks.ModelCheckpoint(\n",
        "    checkpoint_path, \n",
        "    verbose=1, \n",
        "    save_best_only=True,\n",
        "    mode='auto',\n",
        "    save_weights_only=False, \n",
        "    save_freq='epoch'\n",
        ")\n",
        "\n",
        "# 評価値の改善が見られない場合に学習率を減らすコールバック\n",
        "lrp_callback = tf.keras.callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=5, verbose=1)\n",
        "\n",
        "# 早期打ち切り用コールバック\n",
        "es_callback = tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=10, verbose=1) "
      ],
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jV1y-6R63ntD",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "657500d9-1107-4182-db63-aba94cf48e88"
      },
      "source": [
        "epochs = 100\n",
        "\n",
        "history = model.fit(\n",
        "    train_data_gen,\n",
        "    epochs=epochs,\n",
        "    validation_data=validation_data_gen,\n",
        "    callbacks=[cp_callback, lrp_callback, es_callback]\n",
        ")"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "1055/1055 [==============================] - 114s 90ms/step - loss: 0.4584 - accuracy: 0.8532 - val_loss: 0.5188 - val_accuracy: 0.8428\n",
            "\n",
            "Epoch 00001: val_loss improved from inf to 0.51883, saving model to /content/checkpoints/weights.hdf5\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.\n",
            "  category=CustomMaskWarning)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 2/100\n",
            "1055/1055 [==============================] - 92s 87ms/step - loss: 0.2369 - accuracy: 0.9193 - val_loss: 0.2745 - val_accuracy: 0.9098\n",
            "\n",
            "Epoch 00002: val_loss improved from 0.51883 to 0.27449, saving model to /content/checkpoints/weights.hdf5\n",
            "Epoch 3/100\n",
            "1055/1055 [==============================] - 93s 88ms/step - loss: 0.1781 - accuracy: 0.9386 - val_loss: 0.2151 - val_accuracy: 0.9260\n",
            "\n",
            "Epoch 00003: val_loss improved from 0.27449 to 0.21505, saving model to /content/checkpoints/weights.hdf5\n",
            "Epoch 4/100\n",
            "1055/1055 [==============================] - 95s 90ms/step - loss: 0.1436 - accuracy: 0.9497 - val_loss: 0.1838 - val_accuracy: 0.9376\n",
            "\n",
            "Epoch 00004: val_loss improved from 0.21505 to 0.18379, saving model to /content/checkpoints/weights.hdf5\n",
            "Epoch 5/100\n",
            "1055/1055 [==============================] - 94s 89ms/step - loss: 0.1169 - accuracy: 0.9583 - val_loss: 0.2034 - val_accuracy: 0.9332\n",
            "\n",
            "Epoch 00005: val_loss did not improve from 0.18379\n",
            "Epoch 6/100\n",
            "1055/1055 [==============================] - 93s 89ms/step - loss: 0.0972 - accuracy: 0.9656 - val_loss: 0.1735 - val_accuracy: 0.9418\n",
            "\n",
            "Epoch 00006: val_loss improved from 0.18379 to 0.17354, saving model to /content/checkpoints/weights.hdf5\n",
            "Epoch 7/100\n",
            "1055/1055 [==============================] - 94s 89ms/step - loss: 0.0826 - accuracy: 0.9706 - val_loss: 0.1693 - val_accuracy: 0.9423\n",
            "\n",
            "Epoch 00007: val_loss improved from 0.17354 to 0.16935, saving model to /content/checkpoints/weights.hdf5\n",
            "Epoch 8/100\n",
            "1055/1055 [==============================] - 93s 88ms/step - loss: 0.0729 - accuracy: 0.9744 - val_loss: 0.2014 - val_accuracy: 0.9380\n",
            "\n",
            "Epoch 00008: val_loss did not improve from 0.16935\n",
            "Epoch 9/100\n",
            "1055/1055 [==============================] - 93s 88ms/step - loss: 0.0651 - accuracy: 0.9778 - val_loss: 0.2095 - val_accuracy: 0.9411\n",
            "\n",
            "Epoch 00009: val_loss did not improve from 0.16935\n",
            "Epoch 10/100\n",
            "1055/1055 [==============================] - 93s 89ms/step - loss: 0.0583 - accuracy: 0.9798 - val_loss: 0.2110 - val_accuracy: 0.9399\n",
            "\n",
            "Epoch 00010: val_loss did not improve from 0.16935\n",
            "Epoch 11/100\n",
            "1055/1055 [==============================] - 93s 88ms/step - loss: 0.0531 - accuracy: 0.9814 - val_loss: 0.1610 - val_accuracy: 0.9483\n",
            "\n",
            "Epoch 00011: val_loss improved from 0.16935 to 0.16101, saving model to /content/checkpoints/weights.hdf5\n",
            "Epoch 12/100\n",
            "1055/1055 [==============================] - 94s 89ms/step - loss: 0.0481 - accuracy: 0.9834 - val_loss: 0.2024 - val_accuracy: 0.9421\n",
            "\n",
            "Epoch 00012: val_loss did not improve from 0.16101\n",
            "Epoch 13/100\n",
            "1055/1055 [==============================] - 92s 88ms/step - loss: 0.0444 - accuracy: 0.9847 - val_loss: 0.1948 - val_accuracy: 0.9420\n",
            "\n",
            "Epoch 00013: val_loss did not improve from 0.16101\n",
            "Epoch 14/100\n",
            "1055/1055 [==============================] - 93s 88ms/step - loss: 0.0410 - accuracy: 0.9864 - val_loss: 0.1982 - val_accuracy: 0.9436\n",
            "\n",
            "Epoch 00014: val_loss did not improve from 0.16101\n",
            "Epoch 15/100\n",
            "1055/1055 [==============================] - 94s 89ms/step - loss: 0.0374 - accuracy: 0.9871 - val_loss: 0.1769 - val_accuracy: 0.9476\n",
            "\n",
            "Epoch 00015: val_loss did not improve from 0.16101\n",
            "Epoch 16/100\n",
            "1055/1055 [==============================] - 94s 89ms/step - loss: 0.0376 - accuracy: 0.9880 - val_loss: 0.2083 - val_accuracy: 0.9453\n",
            "\n",
            "Epoch 00016: val_loss did not improve from 0.16101\n",
            "\n",
            "Epoch 00016: ReduceLROnPlateau reducing learning rate to 0.0009999999776482583.\n",
            "Epoch 17/100\n",
            "1055/1055 [==============================] - 94s 89ms/step - loss: 0.0262 - accuracy: 0.9922 - val_loss: 0.1749 - val_accuracy: 0.9504\n",
            "\n",
            "Epoch 00017: val_loss did not improve from 0.16101\n",
            "Epoch 18/100\n",
            "1055/1055 [==============================] - 96s 91ms/step - loss: 0.0244 - accuracy: 0.9925 - val_loss: 0.1693 - val_accuracy: 0.9508\n",
            "\n",
            "Epoch 00018: val_loss did not improve from 0.16101\n",
            "Epoch 19/100\n",
            "1055/1055 [==============================] - 95s 90ms/step - loss: 0.0222 - accuracy: 0.9934 - val_loss: 0.1663 - val_accuracy: 0.9520\n",
            "\n",
            "Epoch 00019: val_loss did not improve from 0.16101\n",
            "Epoch 20/100\n",
            "1055/1055 [==============================] - 97s 92ms/step - loss: 0.0209 - accuracy: 0.9938 - val_loss: 0.1692 - val_accuracy: 0.9519\n",
            "\n",
            "Epoch 00020: val_loss did not improve from 0.16101\n",
            "Epoch 21/100\n",
            "1055/1055 [==============================] - 97s 92ms/step - loss: 0.0205 - accuracy: 0.9938 - val_loss: 0.1707 - val_accuracy: 0.9519\n",
            "\n",
            "Epoch 00021: val_loss did not improve from 0.16101\n",
            "\n",
            "Epoch 00021: ReduceLROnPlateau reducing learning rate to 9.999999310821295e-05.\n",
            "Epoch 00021: early stopping\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oksYdmeDnq30"
      },
      "source": [
        "# XXXXXXXX"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gNPuy22_WqqI",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "43d94b5d-b69b-45a1-82be-850254ebc2d1"
      },
      "source": [
        "evaluate_result = model.evaluate_generator(validation_data_gen)\n",
        "print('Validation Loss:' + str(evaluate_result[0]))\n",
        "print('Validation Accuracy:' + str(evaluate_result[1]))"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/training.py:1973: UserWarning: `Model.evaluate_generator` is deprecated and will be removed in a future version. Please use `Model.evaluate`, which supports generators.\n",
            "  warnings.warn('`Model.evaluate_generator` is deprecated and '\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Validation Loss:0.1707179993391037\n",
            "Validation Accuracy:0.9518666863441467\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 287
        },
        "id": "V-A5oXKWm-8Q",
        "outputId": "35272b7a-d412-4c8d-fa79-6cba6cc36175"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "def plot_history(history):\n",
        "    plt.figure(figsize=(19, 6))\n",
        "\n",
        "    # 精度の履歴をプロット\n",
        "    plt.subplot(1, 2, 1)\n",
        "    plt.title('accuracy')\n",
        "    plt.plot(history.history['accuracy'],\"-\",label=\"accuracy\")\n",
        "    plt.plot(history.history['val_accuracy'],\"-\",label=\"val_accuracy\")\n",
        "    plt.xlabel('epoch')\n",
        "    plt.ylabel('accuracy')\n",
        "    plt.legend(loc=\"lower right\")\n",
        " \n",
        "    # 損失の履歴をプロット\n",
        "    plt.subplot(1, 2, 2)\n",
        "    plt.title('loss')\n",
        "    plt.plot(history.history['loss'],\"-\",label=\"loss\",)\n",
        "    plt.plot(history.history['val_loss'],\"-\",label=\"val_loss\")\n",
        "    plt.xlabel('epoch')\n",
        "    plt.ylabel('loss')\n",
        "    plt.legend(loc='upper right')\n",
        "    plt.show()\n",
        "    \n",
        "plot_history(history)"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1368x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vkPqhci35vEm"
      },
      "source": [
        "import math\n",
        "import cv2 as cv\n",
        "import numpy as np\n",
        "\n",
        "def draw_images(images):\n",
        "    column = 6\n",
        "    row = math.ceil(len(images) / column)\n",
        "\n",
        "    plt.figure(figsize=(16, 11))\n",
        "    plt.subplots_adjust(wspace=0.4, hspace=0.6)\n",
        "\n",
        "    for i, image in enumerate(images):\n",
        "        debug_image = cv.imread(image[1])\n",
        "        \n",
        "        plt.subplot(row, column, i+1)\n",
        "        plt.title(str(image[0]), fontsize=10) \n",
        "        plt.tick_params(color='white')\n",
        "        plt.tick_params(labelbottom=False, labelleft=False, labelright=False, labeltop=False)\n",
        "        plt.imshow(cv.cvtColor(debug_image, cv.COLOR_BGR2RGB))\n",
        "        plt.xlabel('', fontsize=15)\n",
        "        plt.ylabel('', rotation=0, fontsize=15, labelpad=20)\n",
        "\n",
        "    plt.show()\n",
        "\n",
        "Y_pred = model.predict_generator(validation_data_gen)\n",
        "y_pred = np.argmax(Y_pred, axis=1)\n",
        "\n",
        "incorrect_numbers = []\n",
        "for true_num, pred_num, filepath in zip(validation_data_gen.classes, y_pred, validation_data_gen.filepaths):\n",
        "    if true_num != pred_num:\n",
        "        if pred_num == 10:\n",
        "            incorrect_numbers.append(['-(10)', filepath])\n",
        "        elif pred_num == 11:\n",
        "            incorrect_numbers.append(['N(11)', filepath])\n",
        "        else:\n",
        "            incorrect_numbers.append([pred_num, filepath])\n",
        "\n",
        "draw_images(incorrect_numbers)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nobuI67-nNIa"
      },
      "source": [
        "# XXXXXXXX"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0WJrVLWjfwlD"
      },
      "source": [
        "model_path = 'checkpoints/weights.hdf5'\n",
        "load_model = tf.keras.models.load_model(model_path)"
      ],
      "execution_count": 29,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7-7-0FsWfzcb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 286
        },
        "outputId": "5180bd00-b49f-4e48-d989-14c852e2f8a6"
      },
      "source": [
        "import cv2\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "test_image = cv2.imread('./dataset/07/00007004.jpg')\n",
        "test_image = cv2.resize(test_image, (96, 96))\n",
        "\n",
        "# グレースケール化\n",
        "test_image = cv2.cvtColor(test_image, cv2.COLOR_BGR2GRAY)\n",
        "\n",
        "# # ヒストグラム平坦化\n",
        "clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))\n",
        "test_image = clahe.apply(test_image)\n",
        "\n",
        "# # ガウシアンブラー\n",
        "test_image = cv2.GaussianBlur(test_image, (3, 3), 0)\n",
        "\n",
        "# 2値化\n",
        "_, test_image = cv2.threshold(test_image, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)\n",
        "\n",
        "test_image = cv2.cvtColor(test_image, cv2.COLOR_GRAY2RGB)\n",
        " \n",
        "plt.imshow(test_image)"
      ],
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f41fe2e2f50>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 33
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "626vf3v-f9ip"
      },
      "source": [
        "from tensorflow.keras.preprocessing.image import img_to_array, load_img\n",
        "\n",
        "# test_image = cv2.resize(test_image, (96, 96))\n",
        "# test_image = cv2.cvtColor(test_image, cv2.COLOR_BGR2RGB)\n",
        "test_image = img_to_array(test_image)\n",
        "test_image = test_image.reshape(-1, 96, 96, 3)\n",
        "test_image = test_image.astype('float32')\n",
        "test_image = test_image * 1.0 / 255"
      ],
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HzWRiagzgErP"
      },
      "source": [
        "predict_result = load_model.predict(test_image)"
      ],
      "execution_count": 35,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BKPp53MGgHbB",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d86499f5-a061-49f8-86bc-415f39cdaf48"
      },
      "source": [
        "print(np.squeeze(predict_result))\n",
        "print(np.argmax(np.squeeze(predict_result)))"
      ],
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[1.2479442e-04 7.1632865e-05 2.4944713e-04 5.0980520e-06 3.6893100e-05\n",
            " 5.7765774e-05 6.1657112e-05 9.9901164e-01 1.9079529e-05 9.9562603e-06\n",
            " 9.1197508e-06 3.4280162e-04]\n",
            "7\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CoKU33ndduuK"
      },
      "source": [
        "# XXXXXXXX"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UhC5NGAsfKNA",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1c4ea3d5-3584-44c0-ac93-614d54fe3c43"
      },
      "source": [
        "load_model.save('7seg_classifier.hdf5', include_optimizer=False)"
      ],
      "execution_count": 37,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.\n",
            "  category=CustomMaskWarning)\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9dMZ3BSAfL0G",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e6b361e0-85be-4363-ac74-9e0695e14026"
      },
      "source": [
        "converter = tf.lite.TFLiteConverter.from_keras_model(model)\n",
        "converter.optimizations = [tf.lite.Optimize.DEFAULT]\n",
        "tflite_quantized_model = converter.convert()\n",
        "\n",
        "open('7seg_classifier.tflite', 'wb').write(tflite_quantized_model)"
      ],
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.\n",
            "  category=CustomMaskWarning)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: /tmp/tmpd76q2ibe/assets\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "2654048"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 38
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "655A6qhXLenW"
      },
      "source": [
        "# XXXXXXXX"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4t0id7YgfQ5b"
      },
      "source": [
        "interpreter = tf.lite.Interpreter(model_path=\"7seg_classifier.tflite\")\n",
        "interpreter.allocate_tensors()"
      ],
      "execution_count": 39,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "I5zdSj-nfSdn",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ba40a5ae-1820-41cc-996f-4ff69fe90295"
      },
      "source": [
        "input_details = interpreter.get_input_details()\n",
        "output_details = interpreter.get_output_details()\n",
        "print(input_details)\n",
        "print(output_details)"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[{'name': 'input_1', 'index': 0, 'shape': array([ 1, 96, 96,  3], dtype=int32), 'shape_signature': array([-1, 96, 96,  3], dtype=int32), 'dtype': <class 'numpy.float32'>, 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array([], dtype=float32), 'zero_points': array([], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}]\n",
            "[{'name': 'Identity', 'index': 177, 'shape': array([ 1, 12], dtype=int32), 'shape_signature': array([-1, 12], dtype=int32), 'dtype': <class 'numpy.float32'>, 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array([], dtype=float32), 'zero_points': array([], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "diQuS7EZfUjZ"
      },
      "source": [
        "interpreter.set_tensor(input_details[0]['index'], test_image)\n",
        "interpreter.invoke()"
      ],
      "execution_count": 41,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "09prFVlpfXBj"
      },
      "source": [
        "tflite_results = interpreter.get_tensor(output_details[0]['index'])"
      ],
      "execution_count": 42,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fy1t99p9fayG",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "20641998-f469-43f8-dd04-5a9a4724f532"
      },
      "source": [
        "print(np.squeeze(tflite_results))\n",
        "print(np.argmax(np.squeeze(tflite_results)))"
      ],
      "execution_count": 43,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[1.5984946e-05 3.2863602e-05 4.4031899e-06 6.6578406e-07 3.5562098e-05\n",
            " 7.7032264e-06 2.6165142e-05 9.9982363e-01 1.2921943e-06 1.3357984e-05\n",
            " 4.9497426e-06 3.3419736e-05]\n",
            "7\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3GnNzgFcAc5j"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}