the code is not right

[swjtulinxi]

def AttentionRefinmentModule(inputs, n_filters):
filters = n_filters

poolingLayer = AveragePooling2D(pool_size=(1, 1), padding='same')？？？？？？

#the paper said it is globalAveragePooling ,but here ?????????

x = poolingLayer(inputs)
x = ConvAndBatch(x, kernel=(1, 1), n_filters=filters, activation='sigmoid')

return multiply([inputs, x])

[MengzhuoTian]

you can set the corresponding pooling size of poolingLayer to achieve a global average pooling.