Python Xception.Trainable示例

def make_finetuned_xcept():

    from keras.applications.xception import Xception
    from keras import layers, optimizers

    xcept = Xception(weights='imagenet', include_top=False, input_shape=(299, 299, 3))

    model = Sequential()
    model.add(xcept)
    model.add(layers.Flatten())
    model.add(layers.Dense(256, use_bias=False))
    model.add(layers.BatchNormalization())
    model.add(layers.Activation("relu"))
    model.add(layers.Dropout(0.5))
    model.add(layers.Dense(2, activation = "softmax"))

    # Unfreeze starting from the last convolution layer

    xcept.Trainable=True
    set_trainable=False
    
    for layer in xcept.layers:
        if layer.name == 'block14_sepconv1':
            set_trainable = True
        if set_trainable:
            layer.trainable = True
        else:
            layer.trainable = False
    
    return model