Fix typo

[matteodg]

Very small fix: channels() is supposed to return -1 when sizes.length!=3, but instead is returning 1.

Out of curiosity is there a reason why in case of sizes.length==2 we do not have rows() and cols() returning respectively size[0] and size[1]?

[saudet]

No reason in particular. I guess that's another bug introduced by @HGuillemet. Please fix it? :)

[HGuillemet]

The 1 instead of -1 is a typo indeed. But at least it doesn't trigger a SIGSEGV like the previous value sizes[2] written by @saudet could :P

For the case sizes.length == 2 I don't know. I don't think there is a logic we could stick to to find proper values for these shortcut methods. I'd vote for deprecating or removing them.
In the general n-dimensional case, the meaning of each dimension could be anything : with or without channel, NHWC/CHWN/NCHW...

[matteodg]

For the case sizes.length == 2 I don't know. I don't think there is a logic we could stick to to find proper values for these shortcut methods. I'd vote for deprecating or removing them.
In the general n-dimensional case, the meaning of each dimension could be anything : with or without channel, NHWC/CHWN/NCHW...

I agree, I vote too to deprecate them as they can be misleading: I was tripped up as well moving from a 2D to a 3D world of coordinates 😄