# Conv layer problems

Hi! I have to simulate a network but I cannot understand this sub-network:

Conv_1 : kernel_size= 3 * 3 * 1 , kernel_number=64
Conv_2…Conv_7: kernel_size= 3 * 3 * 64 , kernel_number=64
Conv_8 : kernel_size= 3 * 3 * 64 , kernel_number=1
what is the In_channels of each layer?
Would you mind helping me figure out this sub network?

I think when you say `kernel_size=a*b*c, kernel_number=d`, a and b are kernel size, c is the number of input channel and d is the number of output channel.

thanks for your reply. It fixed my problem. Could you mind explaining more? I have problems about 3d kernel sizes.

If I understood the question correctly, you want to know more about 3 dimension in kernel size. So for one feature-map that you have, you will have several kernels. For example for your conv_2 layer, you have a kernel with h=3, w=3 and depth of 64 which the same as your feature-map depth. You slide this kernel on your feature-map and do the multiplications and create one feature layer for the feature-map in the next layer. You do this for kernel_number=64 times and create a feature-map of depth 64 for the next layer.