Understanding Conv3d

Can anyone explain the inputs of Conv3d, what Din is which is used as one of its input?
How does it do the convolution?
Any additional explanation to understand how it works is appreciated…

In addition, is there any way to have input in form of: Input: (N,Cin,Din,Hin,Win)
and out put in form of:
(N,Cout,Hout,Wout), so Dout is just one


You should use conv2d then.

What is Din though. is it temporal??

It is just one of the three dimensions the convolution op moves kernel around along. Whether it is time, depth, etc. depends on the user’s interpretation.