Hi @QingyongHu!
Thanks for your work! I noticed that you have adopted a new random sparse annotation approach when working with the KITTI dataset. But I think that approach will result in reselecting sparse labels once per epoch. Shouldn't the index of sparse labels be fixed? Considering the special loading method of the KITTI dataset, is it better to fix the sparse labels directly during the preprocessing?
Hi @QingyongHu! Thanks for your work! I noticed that you have adopted a new random sparse annotation approach when working with the KITTI dataset. But I think that approach will result in reselecting sparse labels once per epoch. Shouldn't the index of sparse labels be fixed? Considering the special loading method of the KITTI dataset, is it better to fix the sparse labels directly during the preprocessing?