我正在尝试使用GoogLeNet的构建石来构建一个简单的初始网络
使用所有4个并行塔楼非常简单,如所有初始描述中所述。但是最后一部分是一个串联,我找不到任何人来描述。
x = Conv2D(64, (1, 1), activation='relu', padding='same')(input)
y = Conv2D(64, (1, 1), activation='relu', padding='same')(input)
y = Conv2D(64, (3, 3), activation='relu', padding='same')(y)
z = Conv2D(64, (1, 1), activation='relu', padding='same')(input)
z = Conv2D(64, (5, 5), activation='relu', padding='same')(z)
u = AveragePooling2D(pool_size=(3, 3), padding='same')(input)
u = Conv2D(64, (1, 1), activation='relu', padding='same' )(u)
但是AveragePooling塔u的输出大小与x,y,z不同。这四个[x,y,z,u]的串联在初始块中如何发生?