mnist_ae.py 文件源码

def decode(y, relu_max):
    assert len(y._keras_shape) == 2
    latent_dim = y._keras_shape[-1]
    x = Reshape((1, 1, latent_dim))(y)
    # 1, 1, latent_dim
    if relu_max:
        x = Activation(utils.scale_up(relu_max))(x)
    # not good? x = BN(mode=2, axis=3)(x)

    batch_size = tf.shape(x)[0]
    x = Deconv2D(40, 7, 7, output_shape=[batch_size, 7, 7, 40], activation='relu',
                 border_mode='same', subsample=(7,7))(x)
    x = BN(mode=2, axis=3)(x)
    # 7, 7, 40
    x = Deconv2D(20, 3, 3, output_shape=[batch_size, 14, 14, 20], activation='relu',
                 border_mode='same', subsample=(2,2))(x)
    x = BN(mode=2, axis=3)(x)
    # 14, 14, 20
    x = Deconv2D(1, 3, 3, output_shape=[batch_size, 28, 28, 1], activation='sigmoid',
                 border_mode='same', subsample=(2,2))(x)
    # 28, 28, 1
    return x