Thanks for your great work. I'm testing other backbones for feature extraction in the lightstereo pipeline. Could you please explain some tips on setting the number of channels?
The channels in your code for mobilenetv2 are set to "channels = [160, 96, 32, 24]". but in fact, I found that the number of channels in Mobilenetv2 is [320, 96, 32, 24, 16]. Are there some reasons about the difference?
If I replace Mobilenetv2 with Mobilenetv3 or Mobilenetv4, how can I set the feature channels?