def FfD(img_channels=1, img_rows=64, img_cols=64):

    input_img = Input(shape=(img_channels, img_rows, img_cols))

    x = Convolution2D(30, 5, 5, init='normal', activation='tanh')(input_img)
    x = ZeroPadding2D((2,2))(x)

    x = Convolution2D(30, 5, 5, init='normal', activation='tanh')(x)
    x = ZeroPadding2D((2,2))(x)
    x = MaxPooling2D((2,2))(x)

    x = Convolution2D(60, 4, 4, init='normal', activation='tanh')(x)

    x = Convolution2D(60, 3, 3, init='normal', activation='tanh')(x)

    x = Convolution2D(120, 3, 3, init='normal', activation='tanh')(x)

    x = Convolution2D(120, 3, 3, init='normal', activation='tanh')(x)

    encoded = Convolution2D(256, 1, 1, init='normal', activation='tanh')(x)

    x = Convolution2D(256, 1, 1, init='normal', activation='tanh')(encoded)

    x = Convolution2D(120, 3, 3, init='normal', activation='tanh')(x)

    x = Convolution2D(120, 3, 3, init='normal', activation='tanh')(x)

    x = Convolution2D(60, 3, 3, init='normal', activation='tanh')(x)

    x = Convolution2D(60, 4, 4, init='normal', activation='tanh')(x)

    x = UpSampling2D((2, 2))(x)
	
    x = Convolution2D(30, 5, 5, init='normal', activation='tanh')(x)

    x = Convolution2D(30, 5, 5, init='normal', activation='tanh')(x)

    x = Flatten()(x)
	
    decoded = Dense(img_channels * img_rows * img_cols, activation='tanh')(x)
    decoded = Reshape((img_channels, img_rows, img_cols))(decoded)

    FfD = Model(input_img, decoded)

    return FfD