Hi, I'm very interested in this paper and try to run it. While I'm using the nuscenes newsplit validation set for eval with the released nusc_newsplit_480_60x30_24e.pth and nusc_newsplit_480_100x50_24e.pth, my evaluation results is below. I wonder if I'm using the right way to eval as the performance is a little bit different from the reported results in paper.
Hi, I'm very interested in this paper and try to run it. While I'm using the nuscenes newsplit validation set for eval with the released nusc_newsplit_480_60x30_24e.pth and nusc_newsplit_480_100x50_24e.pth, my evaluation results is below. I wonder if I'm using the right way to eval as the performance is a little bit different from the reported results in paper.