訓練筆記

一、若是batch過小，訓練的時候不易收斂，loss容易震盪,

二、能夠設置某幾層freeze，不進行參數的更新

for layer in model.layers:

　　layer.trainable = Flase

 

三、

model = Sequential()
model.add(ZeroPadding2D(padding=(1,1),data_format='channels_last',input_shape=(img_width,img_height,channels)))
或者

model = Sequential()

model.add(Dense(32,input_shape=(28,28,1)))

Sequential的第一層，無論是Dense層，仍是padding，須要指定input_shape，注意！這個input_shape不包含有多少條數據，默認shape[0]都是數據的條數

而若是是Model，則須要有input層

inputs = Input(shape=(28,28,1))x = ZeroPadding2D(padding=(1,1))(inputs)x = Conv2D(64,kernel_size=(3,3),activation='relu')(x)x = Conv2D(32,kernel_size=(3,3),activation='relu')(x)x = MaxPooling2D(pool_size=(2,2))(x)x = Dropout(0.25)(x)x = Flatten()(x)x = Dense(128,activation='relu')(x)x = Dropout(0.5)(x)outputs = Dense(num_classes,activation='softmax')(x)