This is a terminology question in the context of ConvNets. So far I've only seen the term "channel" used to refer to the third dimension of input/output volumes. While filters have the same number of dimensions, I haven't seen anybody say for example "this filter has 3 channels." Can we say that? Or do only volumes have channels?
I just read this paper Group Normalization by facebook and saw expressions like
the corresponding channels of these filters can be normalized together, so I believe it makes good sense to use "channels" for the third dimension of a filter in 2DConvNets :)