Fan yidan

about xd-violence datasets, I would like to ask if you have also done the 32-segment processing, and if so, in the same way as your public code？

waht should i do, if i want to get the same per-class result on visda datasets? thanks for your answering