DCGAN Discriminator structure problem

[zzsyjl]

Thank you for sharing this code!

I am confused by the structure of the 5th conv layer of the DCGAN Discriminator.

In your code, the kernel size of this layer is "kernel_size + 1", which is 5. While the input of this layer is of size 6x6.

I think you should set the kernel size to 6, which is corresponding to the first layer of the Generator, where the conv_transpose uses a kernel size of 6.

Or, maybe you have a good reason to deliberately set the kernel size to 5? If so, could you please make an explanation? Thank you very much!