Good job，but a little flaw？
igodogi opened this issue · 1 comments
In “Results on COCO val2017 with detector having human AP of 65.1 on COCO val2017 dataset”： you give the hrnet results just like the original paper rather than the results tested by the same detector of UDP（AP 65.1）? Emma... I am happy to know will the HRNet be better than UDP? Or, what is it that causes the unsatisfactory result of UDP when tested on coco_dev?
Your results:
HRNet results with AP609 person detector:
@igodogi
We did the ablation study in coco-val to show the superiority of UDP over the original HRNet, where we use the same detection result (65.1 AP) for fair comparison. In COCO keypoint test-dev, keypoint results with 65.1 AP detector (ours) have around 0.3~0.4 AP superiority on the counterparts with 60.9 AP detector (original hrnet's). So the comparison in test-dev is not fair. It is just for showcasing the sota performance~