4_pr_dl_lecture.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "MNIST.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "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": "3.7.6"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "hw_h6CwnZ_8B",
        "outputId": "4eaa7e49-163c-4dc3-9e2f-0dac3b5d63cf"
      },
      "source": [
        "import keras\n",
        "import tensorflow as tf\n",
        "from keras.datasets import mnist\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Dropout, Flatten\n",
        "from keras.layers import Conv2D, MaxPooling2D\n",
        "from keras import backend as K\n",
        "batch_size = 128\n",
        "num_classes = 10 # 0～9までの手書き文字\n",
        "epochs = 4 # 訓練データを何回繰り返して学習させるのか\n",
        "\n",
        "img_rows, img_cols = 28, 28\n",
        "\n",
        "# 学習データとテストデータに分割したデータ\n",
        "(x_train, y_train), (x_test, y_test) = mnist.load_data()\n",
        "\n",
        "# backendがTensorflowとTheanoで配列のshapeが異なるために2パターン記述\n",
        "print(K.image_data_format())\n",
        "if K.image_data_format() == 'channels_first':\n",
        "  # 1次元配列に変換\n",
        "  x_train = x_train.reshape(s_train.shape[0], 1, img_rows, img_cols)\n",
        "  x_test = x_test.reshape(x_test.shape[0], 1, img_rows, img_cols)\n",
        "  input_shape = (1, img_rows, img_cols)\n",
        "\n",
        "else:\n",
        "  # 1次元配列に変換\n",
        "  x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1)\n",
        "  x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)\n",
        "  input_shape = (img_rows, img_cols, 1)\n",
        "  \n",
        "# 入力データ[0, 1]の範囲に正規化\n",
        "x_train = x_train.astype('float32')\n",
        "x_test = x_test.astype('float32')\n",
        "# 255で割ったものを新たに変数とする\n",
        "x_train /= 255\n",
        "x_test /= 255\n",
        "\n",
        "print('x_train shape : ', x_train.shape)\n",
        "print(x_train.shape[0], 'train samples')\n",
        "print(x_test.shape[0], 'test samples')\n",
        "\n",
        "# ラベルをOne-Hotベクトルで表現\n",
        "\"\"\"\n",
        "例えば、サンプルに対するターゲットが「5」の場合次のような形になります。\n",
        "[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]\n",
        "[0, 0, 0, 0, 0, 1, 0, 0, 0, 0]\n",
        "\"\"\"\n",
        "y_train = tf.keras.utils.to_categorical(y_train, num_classes)\n",
        "y_test = tf.keras.utils.to_categorical(y_test, num_classes)\n",
        "\n",
        "print(x_train.shape)\n",
        "print(y_train.shape)"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
            "11493376/11490434 [==============================] - 0s 0us/step\n",
            "11501568/11490434 [==============================] - 0s 0us/step\n",
            "channels_last\n",
            "x_train shape :  (60000, 28, 28, 1)\n",
            "60000 train samples\n",
            "10000 test samples\n",
            "(60000, 28, 28, 1)\n",
            "(60000, 10)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "wp5xUGenToW_",
        "outputId": "c0ab6eb9-5795-4ff5-8fff-63e8b199eff6"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rLbg4nIPTVJe",
        "outputId": "c17ccccd-760d-4c57-80a4-298b7ed0d135"
      },
      "source": [
        "# sampling to mak small data for lecture\n",
        "import numpy as np\n",
        "y_train_org = y_train\n",
        "x_train_org = x_train\n",
        "\n",
        "# 60000個から1000個をランダムにサンプリング\n",
        "N = 1000\n",
        "#inds = range(N)\n",
        "inds = np.array(range(60000))\n",
        "np.random.shuffle(inds)\n",
        "inds = inds[:N]\n",
        "y_train = y_train[inds][:]\n",
        "x_train = x_train[inds][:]\n",
        "print(y_train.shape)"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(1000, 10)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "wU2PvVIWS104",
        "outputId": "9d8498b6-6ba6-4cdf-bb82-c3180a21e7b8"
      },
      "source": [
        "# CNNネットワークの構築\n",
        "model = Sequential()\n",
        "model.add(Conv2D(32, kernel_size=(3, 3),\n",
        "                activation='relu',\n",
        "                input_shape=input_shape)) # 3×3のカーネルサイズの2D Convolution layer\n",
        "model.add(Conv2D(64, (3, 3), activation='relu'))\n",
        "model.add(Conv2D(64, (3, 3), activation='relu'))\n",
        "#model.add(Conv2D(64, (3, 3), activation='relu'))\n",
        "#model.add(Conv2D(64, (3, 3), activation='relu'))\n",
        "#model.add(Conv2D(64, (3, 3), activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2, 2))) # max pooling layer\n",
        "model.add(Flatten())\n",
        "model.add(Dense(128, activation='relu')) # 全結合層\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "\n",
        "# 損失関数,最適化関数,評価指標を指定してモデルをコンパイル\n",
        "model.compile(loss=keras.losses.categorical_crossentropy,\n",
        "             #optimizer=keras.optimizers.Adadelta(),\n",
        "             optimizer=tf.keras.optimizers.Adam(),\n",
        "             metrics=['accuracy'])\n",
        "\n",
        "# モデルの学習\n",
        "model.fit(x_train, y_train,\n",
        "         batch_size=batch_size,\n",
        "         epochs=epochs,\n",
        "         verbose=1,\n",
        "         validation_data=(x_test, y_test))"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/4\n",
            "8/8 [==============================] - 32s 270ms/step - loss: 1.6814 - accuracy: 0.5240 - val_loss: 0.7251 - val_accuracy: 0.7685\n",
            "Epoch 2/4\n",
            "8/8 [==============================] - 1s 102ms/step - loss: 0.6560 - accuracy: 0.7960 - val_loss: 0.5189 - val_accuracy: 0.8477\n",
            "Epoch 3/4\n",
            "8/8 [==============================] - 1s 102ms/step - loss: 0.3842 - accuracy: 0.8850 - val_loss: 0.4433 - val_accuracy: 0.8648\n",
            "Epoch 4/4\n",
            "8/8 [==============================] - 1s 101ms/step - loss: 0.3149 - accuracy: 0.8960 - val_loss: 0.3878 - val_accuracy: 0.8799\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7fc94037a810>"
            ]
          },
          "metadata": {},
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 582
        },
        "id": "ybscRk1ghYuX",
        "outputId": "421c87a5-9991-4db1-96c2-e24adc8bb2ba"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "# 学習をグラフ化（正解率）\n",
        "print(model.history.history)\n",
        "acc = model.history.history['accuracy']\n",
        "val_acc = model.history.history['val_accuracy']\n",
        "\n",
        "# Accuracy Plot\n",
        "plt.plot(acc)\n",
        "plt.plot(val_acc)\n",
        "plt.title('Accuracy')\n",
        "plt.legend(['train', 'test'], loc='upper left')\n",
        "plt.show()\n",
        "\n",
        "# Loss Plot\n",
        "loss = model.history.history['loss']\n",
        "val_loss = model.history.history['val_loss']\n",
        "plt.plot(loss ,label = 'training loss')\n",
        "plt.plot(val_loss, label= 'validation loss')\n",
        "plt.title('Training and Validation loss')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "{'loss': [1.6814043521881104, 0.6560326218605042, 0.3842330276966095, 0.3149268627166748], 'accuracy': [0.5239999890327454, 0.7960000038146973, 0.8849999904632568, 0.8960000276565552], 'val_loss': [0.7251490354537964, 0.5189098119735718, 0.44328731298446655, 0.38775312900543213], 'val_accuracy': [0.7684999704360962, 0.8476999998092651, 0.864799976348877, 0.8798999786376953]}\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Un6Osq6cl7sU",
        "outputId": "44c8e2fd-9f23-4ad7-939d-8fae8a5f6f28"
      },
      "source": [
        "# モデルの評価\n",
        "score = model.evaluate(x_test, y_test, verbose=0)\n",
        "print('Test loss:', score[0])\n",
        "print('Test accuracy:', score[1])"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Test loss: 0.3877529203891754\n",
            "Test accuracy: 0.8798999786376953\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "drm6LLW4dN4H",
        "outputId": "3ae120ed-e19b-4c86-d030-fbdde78bd269"
      },
      "source": [
        "# モデルの概要を表示\n",
        "model.summary()\n",
        "    \n",
        "# 概要を画像に出力\n",
        "#plot_model(model, to_file='model.png')"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d (Conv2D)              (None, 26, 26, 32)        320       \n",
            "_________________________________________________________________\n",
            "conv2d_1 (Conv2D)            (None, 24, 24, 64)        18496     \n",
            "_________________________________________________________________\n",
            "conv2d_2 (Conv2D)            (None, 22, 22, 64)        36928     \n",
            "_________________________________________________________________\n",
            "max_pooling2d (MaxPooling2D) (None, 11, 11, 64)        0         \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 7744)              0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 128)               991360    \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 10)                1290      \n",
            "=================================================================\n",
            "Total params: 1,048,394\n",
            "Trainable params: 1,048,394\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 349
        },
        "id": "Br9g2kiRszY_",
        "outputId": "76b49c9d-3db9-4bcd-f5a2-59e12998d604"
      },
      "source": [
        "# 1つのサンプルデータを用いてテスト\n",
        "from skimage import io\n",
        "import numpy as np\n",
        "import random\n",
        "# test dataをランダムに取得\n",
        "ind = np.int16(np.round(random.random()*x_test.shape[0]))\n",
        "print(ind)\n",
        "x_sample = x_test[ind]\n",
        "y_sample = y_test[ind]\n",
        "\n",
        "# 画像を可視化\n",
        "x_sample2 = x_sample.reshape(x_sample.shape[0], x_sample.shape[1])\n",
        "print(y_sample)\n",
        "io.imshow(x_sample2)\n",
        "#plt.show()\n",
        "\n",
        "# 予測関数入力用に変形\n",
        "x_sample = x_sample.reshape(1, img_rows, img_cols, 1)\n",
        "\n",
        "# ネットワークによる予測\n",
        "expect = model.predict(x_sample)\n",
        "print(np.round(expect))"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "6373\n",
            "[0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]\n",
            "[[0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]]\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}