ramen_hayashi.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "ramen_hayashi.ipynb",
      "provenance": [],
      "collapsed_sections": [
        "MO1WB3Jy-nGw"
      ],
      "mount_file_id": "15zubguWAN_l3xsQa377LOrNDNRHKqOZp",
      "authorship_tag": "ABX9TyMmbegAbZFuvSAWk5nZ6i6+",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/snsk/Android_WariCaLoc/blob/master/ramen_hayashi.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "cwC1HfVnohpT",
        "outputId": "a960c28b-8e1b-4502-9b6f-2eca801666f1"
      },
      "source": [
        "import cv2\n",
        "import numpy as np\n",
        "from google.colab.patches import cv2_imshow\n",
        "\n",
        "colab_root = '/content/drive/MyDrive/colab_root/'\n",
        "metamorphic_images_store = '/content/drive/MyDrive/colab_root/metamorphic_images/'\n",
        "src_image_filname = 'hiyashi_3.jpg'\n",
        "\n",
        "img = cv2.imread(colab_root + src_image_filname,1)\n",
        "rows,cols = img.shape[:-1]\n",
        "\n",
        "for i in range(1, 100):\n",
        "  #画像を回転させて背景を白で塗りつぶす\n",
        "  M = np.float32([[1,0,i * 10],[0,1,0]])\n",
        "  dst = cv2.warpAffine(img,M,(cols,rows),borderValue=(255, 255, 255))\n",
        "  #cv2_imshow(dst)\n",
        "  cv2.imwrite(metamorphic_images_store + 'T'+ str(i*10) + '_'+ src_image_filname, dst)\n",
        "\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "from keras.preprocessing import image\n",
        "from keras.models import load_model\n",
        "from keras.preprocessing.image import img_to_array, load_img\n",
        "import pprint\n",
        "\n",
        "# modelの読み込み\n",
        "model = load_model(colab_root + 'ramen_hiyashi_acc0.9675.h5')\n",
        "# 正解ラベルの定義\n",
        "label=['ramen', 'hiyashi']\n",
        "x=[]\n",
        "y=[]\n",
        "\n",
        "for i in range(1, 100):\n",
        "  # 画像を読み込む。\n",
        "  img = image.load_img(metamorphic_images_store + 'T'+ str(i*10) + '_'+ src_image_filname, target_size=(100,100))\n",
        "  img_nad = img_to_array(img)/255\n",
        "  img_nad = img_nad[None, ...]\n",
        "\n",
        "  # 判別\n",
        "  pred = model.predict(img_nad, batch_size=1, verbose=0)\n",
        "  score = np.max(pred)\n",
        "  pred_label = label[np.argmax(pred[0])]\n",
        "  #print('src:R',str(i*10),'name:',pred_label,'score:',score)\n",
        "  if pred_label == 'ramen': #誤判定の場合はスコアを0に\n",
        "    score = 0\n",
        "  x.append(i*10)\n",
        "  y.append(score)\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "plt.plot(x,y,label=\"acc\")\n",
        "plt.title('test result - translation')\n",
        "plt.xlabel('translation x value')\n",
        "plt.ylabel('accuracy')\n",
        "plt.grid(True)\n",
        "plt.legend()\n",
        "plt.show()\n"
      ],
      "execution_count": 45,
      "outputs": [
        {
          "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": "markdown",
      "metadata": {
        "id": "uLKR6_efYN8o"
      },
      "source": [
        "# メタモルフィックテスティング:回転"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "rftzfi_jYbL0",
        "outputId": "a24092af-f479-47ac-c703-f9bb03ca317f"
      },
      "source": [
        "import cv2\n",
        "from google.colab.patches import cv2_imshow\n",
        "\n",
        "colab_root = '/content/drive/MyDrive/colab_root/'\n",
        "metamorphic_images_store = '/content/drive/MyDrive/colab_root/metamorphic_images/'\n",
        "src_image_filname = 'hiyashi_3.jpg'\n",
        "\n",
        "img = cv2.imread(colab_root + src_image_filname,1)\n",
        "rows,cols = img.shape[:-1]\n",
        "\n",
        "for i in range(1, 37):\n",
        "  #画像を回転させて背景を白で塗りつぶす\n",
        "  M = cv2.getRotationMatrix2D((cols/2,rows/2), i * 10, 1)\n",
        "  dst = cv2.warpAffine(img,M,(cols,rows),borderValue=(255, 255, 255))\n",
        "  #cv2_imshow(dst)\n",
        "  cv2.imwrite(metamorphic_images_store + 'R'+ str(i*10) + '_'+ src_image_filname, dst)\n",
        "\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "from keras.preprocessing import image\n",
        "from keras.models import load_model\n",
        "from keras.preprocessing.image import img_to_array, load_img\n",
        "import pprint\n",
        "\n",
        "# modelの読み込み\n",
        "model = load_model(colab_root + 'ramen_hiyashi_acc0.9675.h5')\n",
        "# 正解ラベルの定義\n",
        "label=['ramen', 'hiyashi']\n",
        "x=[]\n",
        "y=[]\n",
        "\n",
        "for i in range(1, 37):\n",
        "  # 画像を読み込む。\n",
        "  img = image.load_img(metamorphic_images_store + 'R'+ str(i*10) + '_'+ src_image_filname, target_size=(100,100))\n",
        "  img_nad = img_to_array(img)/255\n",
        "  img_nad = img_nad[None, ...]\n",
        "\n",
        "  # 判別\n",
        "  pred = model.predict(img_nad, batch_size=1, verbose=0)\n",
        "  score = np.max(pred)\n",
        "  pred_label = label[np.argmax(pred[0])]\n",
        "  #print('src:R',str(i*10),'name:',pred_label,'score:',score)\n",
        "  if pred_label == 'ramen': #誤判定の場合はスコアを0に\n",
        "    score = 0\n",
        "  x.append(i*10)\n",
        "  y.append(score)\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "plt.plot(x,y,label=\"acc\")\n",
        "plt.title('test result - rotation')\n",
        "plt.xlabel('degree of rotation')\n",
        "plt.ylabel('accuracy')\n",
        "plt.grid(True)\n",
        "plt.legend()\n",
        "plt.show()\n"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "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": "markdown",
      "metadata": {
        "id": "MO1WB3Jy-nGw"
      },
      "source": [
        "# 画像1枚から回転360度=36枚を作成して保存"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Rws0RyCB-219"
      },
      "source": [
        "import cv2\n",
        "from google.colab.patches import cv2_imshow\n",
        "\n",
        "colab_root = '/content/drive/MyDrive/colab_root/'\n",
        "metamorphic_images_store = '/content/drive/MyDrive/colab_root/metamorphic_images/'\n",
        "src_image_filname = 'hiyashi_3.jpg'\n",
        "\n",
        "img = cv2.imread(colab_root + src_image_filname,1)\n",
        "rows,cols = img.shape[:-1]\n",
        "\n",
        "for i in range(1, 5):\n",
        "  #画像を回転させて背景を白で塗りつぶす\n",
        "  M = cv2.getRotationMatrix2D((cols/2,rows/2), i * 10, 1)\n",
        "  dst = cv2.warpAffine(img,M,(cols,rows),borderValue=(255, 255, 255))\n",
        "  cv2_imshow(dst)\n",
        "  cv2.imwrite(metamorphic_images_store + 'R'+ str(i*10) + '_'+ src_image_filname, dst)\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YjEMSfl-6vK_"
      },
      "source": [
        "# 画像変形した1枚を判定"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GtzlmVc960kM",
        "outputId": "9616b127-de52-47a2-bafb-d4468efa6450"
      },
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "from keras.preprocessing import image\n",
        "from keras.models import load_model\n",
        "from keras.preprocessing.image import img_to_array, load_img\n",
        "import pprint\n",
        "\n",
        "colab_root = '/content/drive/MyDrive/colab_root/'\n",
        "\n",
        "# 画像を読み込む。\n",
        "img = image.load_img(colab_root+'hiyashi_3.jpg',  target_size=(100,100))\n",
        "img = np.array(img)\n",
        "\n",
        "#plt.imshow(img)\n",
        "#plt.show()\n",
        "\n",
        "# flow に渡すために配列を四次元化\n",
        "img = img[None, ...]\n",
        "\n",
        "# 画像変形のジェネレータを作成\n",
        "datagen = image.ImageDataGenerator(rotation_range=0)\n",
        "gen = datagen.flow(img, batch_size = 1)\n",
        "\n",
        "# バッチの実行\n",
        "batches = next(gen)\n",
        "g_img = batches[0]/255\n",
        "g_img = g_img[None, ...]\n",
        "\n",
        "# modelの読み込み\n",
        "model = load_model(colab_root + 'ramen_hiyashi_acc0.9675.h5')\n",
        "# 正解ラベルの定義\n",
        "label=['ramen', 'hiyashi']\n",
        "\n",
        "# 判別\n",
        "pred = model.predict(g_img, batch_size=1, verbose=0)\n",
        "score = np.max(pred)\n",
        "pred_label = label[np.argmax(pred[0])]\n",
        "print('name:',pred_label)\n",
        "print('score:',score)\n"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "name: hiyashi\n",
            "score: 0.9987631\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "l5yo8NS6ECqT"
      },
      "source": [
        "# keras の画像変形を試す\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vtey08KRECWj"
      },
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "from keras.preprocessing import image\n",
        "\n",
        "colab_root = '/content/drive/MyDrive/colab_root/'\n",
        "\n",
        "# 画像を読み込む。\n",
        "img = image.load_img(colab_root+'ramen_1.jpg')\n",
        "img = np.array(img)\n",
        "\n",
        "plt.imshow(img)\n",
        "plt.show()\n",
        "\n",
        "# flow に渡すために配列を四次元化\n",
        "img = img[None, ...]\n",
        "\n",
        "# 画像変形のジェネレータを作成\n",
        "datagen = image.ImageDataGenerator(rotation_range=50)\n",
        "gen = datagen.flow(img, batch_size = 1)\n",
        "\n",
        "# バッチの実行\n",
        "batches = next(gen)\n",
        "# 画像として表示するためにuint8に変換\n",
        "g_img = batches[0].astype(np.uint8)\n",
        "\n",
        "plt.imshow(g_img)\n",
        "plt.show()\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GOcH73LJ-Ca9"
      },
      "source": [
        "#2種類の画像を識別するモデル"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JZOAZidS6GVy"
      },
      "source": [
        "import keras\n",
        "from keras.utils.vis_utils import plot_model\n",
        "from keras.utils import np_utils\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Dropout, Activation\n",
        "from keras.optimizers import SGD\n",
        "from keras.models import Sequential\n",
        "from keras.layers.convolutional import Conv2D, MaxPooling2D\n",
        "from keras.layers.core import Dense, Dropout, Activation, Flatten\n",
        "import numpy as np\n",
        "from sklearn.model_selection import train_test_split\n",
        "from PIL import Image\n",
        "import glob\n",
        "\n",
        "folder = [\"currypan\",\"piroshik\"]\n",
        "image_size = 100\n",
        "X = []\n",
        "Y = []\n",
        "for index, name in enumerate(folder):\n",
        "    dir = \"/content/drive/MyDrive/colab_root/\" + name\n",
        "    files = glob.glob(dir + \"/*.jpg\")\n",
        "    for i, file in enumerate(files):\n",
        "        image = Image.open(file)\n",
        "        image = image.convert(\"RGB\")\n",
        "        image = image.resize((image_size, image_size))\n",
        "        data = np.asarray(image)\n",
        "        X.append(data)\n",
        "        Y.append(index)\n",
        "\n",
        "X = np.array(X)\n",
        "Y = np.array(Y)\n",
        "\n",
        "X = X.astype('float32')\n",
        "X = X / 255.0\n",
        "\n",
        "# 正解ラベルの形式を変換\n",
        "Y = np_utils.to_categorical(Y, 2)\n",
        "\n",
        "# 学習用データとテストデータ\n",
        "X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.20)\n",
        "'''\n",
        "X_train:訓練用入力データ\n",
        "X_test :訓練用正解データ\n",
        "Y_train:評価用入力データ\n",
        "Y_test :評価用正解データ\n",
        "'''\n",
        "\n",
        "# CNNを構築\n",
        "# Sequentialモデルは addされたモデル層が一列に並ぶ\n",
        "model = Sequential()\n",
        "\n",
        "#Conv2D:2次元畳み込み層 3x3のパネルを32枚使う。padding='same' で出力画像のサイズが変わらないようにする\n",
        "model.add(Conv2D(32, (3, 3), padding='same',input_shape=X_train.shape[1:], activation='relu'))\n",
        "model.add(Conv2D(32, (3, 3), activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "model.add(Conv2D(64, (3, 3), padding='same', activation='relu'))\n",
        "model.add(Conv2D(64, (3, 3), activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "model.add(Dropout(0.5))\n",
        "\n",
        "model.add(Flatten())\n",
        "model.add(Dense(512, activation='relu'))\n",
        "model.add(Dropout(0.5))\n",
        "model.add(Dense(2, activation='softmax'))\n",
        "\n",
        "#モデルのコンパイル\n",
        "model.compile(\n",
        "    loss='categorical_crossentropy',\n",
        "    optimizer='sgd',\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "\n",
        "#モデルの訓練\n",
        "epochs_num=500\n",
        "result = model.fit(X_train, Y_train, epochs=epochs_num, validation_data=(X_test, Y_test))\n",
        "\n",
        "#modelのビジュアル可視化。要install and PATH setting GraphViz, pydot\n",
        "#functional API を利用した複雑なモデルの時はこちら\n",
        "plot_model(model, to_file='./model.png', show_shapes=True, expand_nested=True)\n",
        "#modelのテキスト可視化。Sequentialモデルならこれでも十分、らしい\n",
        "#print(model.summary())\n",
        "\n",
        "#model.save('ramen_hiyashi.h5')\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        " \n",
        "plt.plot(range(1, epochs_num+1), result.history['accuracy'], label=\"training\")\n",
        "plt.plot(range(1, epochs_num+1), result.history['val_accuracy'], label=\"validation\")\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.legend()\n",
        "plt.show()\n",
        "\n",
        "#評価 & 評価結果出力\n",
        "print(model.evaluate(X_test, Y_test))\n",
        "print('Test:')\n",
        "test_loss, test_acc = model.evaluate(X_test, Y_test, verbose=1)\n",
        "print('\\nTest accuracy:', test_acc)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MhYxElk3dQYt"
      },
      "source": [
        "# 構築済みのモデルを使う\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "suH0mx8pdPsM",
        "outputId": "dd9b73f5-1554-46fe-da63-ebe1113dfc05"
      },
      "source": [
        "from keras.models import load_model\n",
        "import numpy as np\n",
        "from keras.preprocessing.image import img_to_array, load_img\n",
        "\n",
        "colab_root = '/content/drive/MyDrive/colab_root/'\n",
        "\n",
        "# modelと画像ファイルの読み込み\n",
        "model = load_model(colab_root + 'ramen_hiyashi_acc0.9675.h5')\n",
        "\n",
        "# 画像ファイルの行列変換\n",
        "img = img_to_array(load_img(colab_root + 'ramen_2.jpg', target_size=(100,100)))\n",
        "img_nad = img_to_array(img)/255\n",
        "img_nad = img_nad[None, ...]\n",
        "\n",
        "# 正解ラベルの定義\n",
        "label=['ramen', 'hiyashi']\n",
        "\n",
        "# 判別\n",
        "pred = model.predict(img_nad, batch_size=1, verbose=0)\n",
        "print(pred)\n",
        "score = np.max(pred)\n",
        "pred_label = label[np.argmax(pred[0])]\n",
        "print('name:',pred_label)\n",
        "print('score:',score)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[0.7606613  0.23933864]]\n",
            "name: ramen\n",
            "score: 0.7606613\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ebw76r-VdeXP"
      },
      "source": [
        "\n",
        "# np_arrayとimg_nad の動作確認"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Vc8ncqV8GWxL"
      },
      "source": [
        "from keras.models import load_model\n",
        "import numpy as np\n",
        "from keras.preprocessing.image import img_to_array, load_img\n",
        "import pprint\n",
        "\n",
        "colab_root = '/content/drive/MyDrive/colab_root/'\n",
        "\n",
        "# modelと画像ファイルの読み込み\n",
        "model = load_model(colab_root + 'ramen_hiyashi_acc0.9675.h5')\n",
        "\n",
        "# 画像ファイルの行列変換\n",
        "img = img_to_array(load_img(colab_root + 'ramen_2.jpg', target_size=(100,100)))\n",
        "img_nad = img_to_array(img)/255\n",
        "img_nad = img_nad[None, ...]\n",
        "pprint.pprint(img_nad)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "m2IaBmxRB34A"
      },
      "source": [
        "# colab sandbox"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ytiPvLKWB9jK",
        "outputId": "ebce55b2-dab4-42c5-c561-f974796aa60a"
      },
      "source": [
        "!python -V"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Python 3.7.11\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}