如何在Keras中的卷积层之间添加跳过连接

时间:2019-02-21 14:39:41

标签: python keras conv-neural-network deep-residual-networks

我想在keras中的剩余块之间添加一个跳过连接。这是我当前的实现,因为张量具有不同的形状,因此不起作用。

函数如下:

def build_res_blocks(net, x_in, num_res_blocks, res_block, num_filters, res_block_expansion, kernel_size, scaling):
    net_next_in = net
    for i in range(num_res_blocks):
        net = res_block(net_next_in, num_filters, res_block_expansion, kernel_size, scaling)

        # net tensor shape: (None, None, 32)
        # x_in tensor shape: (None, None, 3)
        # Error here, net_next_in should be in the shape of (None, None, 32) to be fed into next layer
        net_next_in = Add()([net, x_in]) 

    return net

我得到的错误是:ValueError: Operands could not be broadcast together with shapes (None, None, 32) (None, None, 3)

我的问题是,如何将这些张量添加或合并为正确的形状(无,无,32)。如果这不是正确的方法,那么您如何才能达到预期的效果?

编辑:

这是res_block的样子:

def res_block(x_in, num_filters, expansion, kernel_size, scaling):
    x = Conv2D(num_filters * expansion, kernel_size, padding='same')(x_in)
    x = Activation('relu')(x)
    x = Conv2D(num_filters, kernel_size, padding='same')(x)
    x = Add()([x_in, x])
return x

1 个答案:

答案 0 :(得分:2)

您不能添加不同形状的张量。您可以将它们与keras.layers.Concatenate串联在一起,但这会使您的形状为[None, None, 35]的张量。

或者,看看 Resnet50在Keras中的实现。对于要添加的尺寸不同的情况,它们的残差块在快捷方式中具有1x1xC卷积。