About AdaFocus+

[dreamerlin]

Thanks for your great work!

I want to know which part of the codes is about AdaFouces+ since I wonder how to batch-process videos with different number of frames for some skipped ones. Or does it process per frame in a video one by one?

[dreamerlin]

maybe it is in models/gfv_net/one_step_act

I checked it but it seems not AdaFocus+

[FrozenBurning]

Thanks for your interest in our work! However, this repo does not contain the code of AdaFocus+. We are working on it and will update the code soon. As for the implementation of AdaFocus+, it process per frame in a video one by one (similar to stage 2 of AdaFocus). Building on top of TSM, we only use AdaFocus which has already achieved pretty good performance in terms of accuracy and inference speed. Thus, we leave AdaFocus+ on TSM for future works.