How to apply CAI when different layers have varying dimensional scales, and ultimately merge the computation results?

[achinastone]

[achinastone]

By the way, Suggestion:‌ The ‌NormalizeMax‌ of ‌TNNetVolume‌ can be normalized by different depths, because the ‌dimensions may vary‌

[joaopauloschuler]

Hello and good morning - can you clarify please what do you mean by "varying dimensional scales"? Should I understand that you mean "the same X and Y with different depth"? If yes, you can concatenate with TNNetDeepConcat.

[achinastone]

"NormalizeMax" refers to taking the maximum value across all flattened data dimensions. I think it might be necessary to apply NormalizeMax separately on different channels. Of course, maybe I'm just inexperienced and don't fully understand many concepts.

[joaopauloschuler]

I think that you are looking for TNNetChannelStdNormalization.

[achinastone]

Oh,thanks