深度學習 - Kaggle - Dog & Cat - 不做任何處理直接訓練

深度學習 - Kaggle - Dog & Cat - 不做任何處理直接訓練

#資料夾分類處理後的路徑
original_dataset_dir ='C:\\Users\\lido_lee\\Downloads\\kaggle_original_data'
base_dir = 'C:\\Users\\lido_lee\\Downloads\\cats_and_dogs_small'
train_dir = 'C:\\Users\\lido_lee\\Downloads\\cats_and_dogs_small\\train'
validation_dir = 'C:\\Users\\lido_lee\\Downloads\\cats_and_dogs_small\\validation'

將Kaggle下載的資料分類處理，之後再來複習，這邊先跳過詳細過程。

#建立模型網絡
from keras import layers
from keras import models
model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu',input_shape=(150, 150, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))

#選擇優化器RMSprop、損失函數binary_crossentropy、指標acc
from keras import optimizers
model.compile(loss='binary_crossentropy', 
              optimizer=optimizers.RMSprop(lr=1e-4), 
              metrics=['acc'])

#使用Keras內建圖型Data生成器
from keras.preprocessing.image import ImageDataGenerator
train_datagen = ImageDataGenerator(rescale=1./255)
test_datagen = ImageDataGenerator(rescale=1./255)

若要使用圖形增強，直接使用ImageDataGenerator的function即可，後續會用到。

#將訓練資料縮小至150*150，每個批量20張圖，所以2000張圖跌代一次要跑100次
train_generator = train_datagen.flow_from_directory(train_dir,
                                                    target_size=(150, 150),
                                                    batch_size=20,
                                                    class_mode='binary')

#驗證資料同上
validation_generator = test_datagen.flow_from_directory( validation_dir,
                                                         target_size=(150, 150),
                                                         batch_size=20,
                                                         class_mode='binary')

#將訓練&驗證資料丟進fit_generator訓練
history = model.fit_generator( train_generator,
                               steps_per_epoch=100,
                               epochs=30,
                               validation_data=validation_generator,
                               validation_steps=50)

#將訓練完的模型&參數儲存起來
model.save('cats_and_dogs_small_1.h5')