DRNet icon indicating copy to clipboard operation
DRNet copied to clipboard
verified publisher icon taohan10200

missing label for SENSE dataset

Open linthmwan opened this issue 3 years ago • 6 comments

I download the SENSE dataset from baidu, but there is no label for scene IMG_4696_cut_03.txt at frame 000084.jpg and 000085.jpg. Can anyone provide GT of these two frames? thanks!

linthmwan avatar Feb 10 '23 13:02 linthmwan

  • ...
  • 000082.jpg 1280 720 511.866 4.628 527.381 18.839 521.197 9.768 13
  • 000083.jpg 1280 720 524.209 0 538.7 12.031 532.011 4.131 13
  • 000084.jpg 1280 720
  • 000085.jpg 1280 720
  • ...

Thanks for your reminder, it just because there is no object in these two frames.

taohan10200 avatar Feb 10 '23 17:02 taohan10200

IMG_4696_cut_03.txt Above is the file I download, I also checked that there are people in the frames 84.jpg and 85.jpg exactly.

linthmwan avatar Feb 10 '23 17:02 linthmwan

My file is the same with yours, I guess the annotators miss some objects in the whole dataset.

taohan10200 avatar Feb 10 '23 17:02 taohan10200

Thanks for reply! I found this issue because of running the test_SENSE.py and it will have error at VIC.py line 206 match_gt, pois = self.get_ROI_and_MatchInfo(target[0], target[1],noise=None) and successively in points_from_den.py unmatched0 = torch.where(dis.min(1)[0]>0)[0] because there is no target. Is there a correct way to run the test_SENSE.py?

linthmwan avatar Feb 10 '23 18:02 linthmwan

The simple way is to skip this scene. We will try to fix it later.

taohan10200 avatar Feb 11 '23 08:02 taohan10200

Got it. Thank you for your patience.

linthmwan avatar Feb 11 '23 08:02 linthmwan