引き続き写経して心を落ち着けております。エルベです。
snack.elve.club
www.tensorflow.org
一応動いて、判別もできたんで学習結果だけダウンロードしてローカルで実行しようとしてはまってます。
google Pythonのバージョン3.7.12だったんだもん。
ワイのところはもう3.10～だったからさー(ﾉД`)ﾅｪﾙｼｸ もう開発一旦止まって！！（タイトル回収）
とりあえず現在動くコード
fit_generatorは使えなくなる予定らしいからなんとかしないとね(^_^;)
import tensorflow as tf from tensorflow.keras.models import Sequential from tensorflow.keras.layers import Dense, Conv2D, Flatten, Dropout, MaxPooling2D from tensorflow.keras.preprocessing.image import ImageDataGenerator from tensorflow.keras.preprocessing.image import img_to_array, load_img import os import numpy as np import matplotlib.pyplot as plt train_dir = os.path.join("./drive/MyDrive/img/", 'train') validation_dir = os.path.join("./drive/MyDrive/img/", 'validation') train_pic_dir = os.path.join(train_dir, '学習用ピクミン画像') # 学習用ピクミン画像 train_etc_dir = os.path.join(train_dir, '学習用その他') # 学習用 validation_pic_dir = os.path.join(validation_dir, 'ピクミン') # 検証用のピクミン validation_etc_dir = os.path.join(validation_dir, 'その他') # 検証用 num_pic_tr = len(os.listdir(train_pic_dir)) num_etc_tr = len(os.listdir(train_etc_dir)) num_pic_val = len(os.listdir(validation_pic_dir)) num_etc_val = len(os.listdir(validation_etc_dir)) total_train = num_pic_tr + num_etc_tr total_val = num_pic_val + num_etc_val print('total training cat images:', num_pic_tr) print('total training dog images:', num_etc_tr) print('total validation cat images:', num_pic_val) print('total validation dog images:', num_etc_val) print("--") print("Total training images:", total_train) print("Total validation images:", total_val) batch_size = 128 epochs = 15 IMG_HEIGHT = 150 IMG_WIDTH = 150 train_image_generator = ImageDataGenerator(rescale=1./255) # 学習データのジェネレータ validation_image_generator = ImageDataGenerator(rescale=1./255) # 検証データのジェネレータ train_data_gen = train_image_generator.flow_from_directory(batch_size=batch_size, directory=train_dir, shuffle=True, target_size=(IMG_HEIGHT, IMG_WIDTH), class_mode='binary') val_data_gen = validation_image_generator.flow_from_directory(batch_size=batch_size, directory=validation_dir, target_size=(IMG_HEIGHT, IMG_WIDTH), class_mode='binary') sample_training_images, _ = next(train_data_gen) # この関数は、1行5列のグリッド形式で画像をプロットし、画像は各列に配置されます。 def plotImages(images_arr): fig, axes = plt.subplots(1, 5, figsize=(20,20)) axes = axes.flatten() for img, ax in zip( images_arr, axes): ax.imshow(img) ax.axis('off') plt.tight_layout() plt.show() plotImages(sample_training_images[:5]) model = Sequential([ Conv2D(16, 3, padding='same', activation='relu', input_shape=(IMG_HEIGHT, IMG_WIDTH ,3)), MaxPooling2D(), Conv2D(32, 3, padding='same', activation='relu'), MaxPooling2D(), Conv2D(64, 3, padding='same', activation='relu'), MaxPooling2D(), Flatten(), Dense(512, activation='relu'), Dense(1, activation='sigmoid') ]) model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy']) model.summary() history = model.fit_generator( train_data_gen, steps_per_epoch=total_train // batch_size, epochs=epochs, validation_data=val_data_gen, validation_steps=num_validation // batch_size ) acc = history.history['accuracy'] val_acc = history.history['val_accuracy'] loss = history.history['loss'] val_loss = history.history['val_loss'] epochs_range = range(epochs) plt.figure(figsize=(8, 8)) plt.subplot(1, 2, 1) plt.plot(epochs_range, acc, label='Training Accuracy') plt.plot(epochs_range, val_acc, label='Validation Accuracy') plt.legend(loc='lower right') plt.title('Training and Validation Accuracy') plt.subplot(1, 2, 2) plt.plot(epochs_range, loss, label='Training Loss') plt.plot(epochs_range, val_loss, label='Validation Loss') plt.legend(loc='upper right') plt.title('Training and Validation Loss') plt.show()
ここで
こんな感じ。別に調整不要そうだけど写経を続ける（学習に反映させる前にハマった）
#水平反転 image_gen = ImageDataGenerator(rescale=1./255, horizontal_flip=True) train_data_gen = image_gen.flow_from_directory(batch_size=batch_size, directory=train_dir, shuffle=True, target_size=(IMG_HEIGHT, IMG_WIDTH)) augmented_images = [train_data_gen[0][0][0] for i in range(5)] # 上で学習用画像の可視化のために定義、使用されたおなじカスタムプロット関数を再利用する plotImages(augmented_images) #ランダムに左右45度の範囲で回転 image_gen = ImageDataGenerator(rescale=1./255, rotation_range=45) train_data_gen = image_gen.flow_from_directory(batch_size=batch_size, directory=train_dir, shuffle=True, target_size=(IMG_HEIGHT, IMG_WIDTH)) augmented_images = [train_data_gen[0][0][0] for i in range(5)] plotImages(augmented_images)
ここで飽きて、学習結果を保存し、次回は読み込むように仕様、と画策
p='./drive/MyDrive/' model_json_str = model.to_json() open(p+'model.json', 'w').write(model_json_str) model.save_weights(p+'weights.h5');
試しにピクミンの画像を判定
## 次回モデル読み込み # from tensorflow.keras.models import model_from_json # from tensorflow.keras.preprocessing.image import img_to_array, load_img # p = './drive/MyDrive/' # print(p+'model.json') ## モデルを読み込む # model = model_from_json(open(p+'model.json').read()) ## 学習結果を読み込む # model.load_weights(p + 'weights.h5') img_path = os.path.join(validation_dir, '20220113_115516000_iOS.png') print(img_path) img = img_to_array(load_img(img_path, target_size=(150, 150))) img_nad = img_to_array(img)/255 img_nad = img_nad[None, ...] label = ['pic', 'etc'] pred = model.predict(img_nad, batch_size=1, verbose=0) pred_label = label[np.argmax(pred[0])] if pred_label == 'pic': print("pic!!")
無事9割以上の確率でピクミンだと検出。よしっ！
