search

jin-s13 / COCO-WholeBody

ECCV2020 paper "Whole-Body Human Pose Estimation in the Wild"
767 stars 73 forks source link

Foot keypoint sigmas #11

Closed DuncanZauss closed 3 years ago

DuncanZauss commented 3 years ago

Hello! Thank you for making this nice work available. In myeval_foot.py, line 163 you use the following sigmas: sigmas = np.array([0.68, 0.66, 0.66, 0.92, 0.94, 0.94]) / 10.0 So you are using different sigmas for the left and right foot. Could you let me know, why this is the case? All the other sigmas are symmetrical, so that the corresponding left and right body parts have the same sigma. Best, Duncan

jin-s13 commented 3 years ago

The sigmas are calculated individually/separately for each keypoint. So it is possible that some symmetric parts have different sigmas. For the face keypoints, the sigmas are not symmetric either. For hand keypoints, we treat the left and the right hands equally (for model design, and evaluation). So we specially average the sigmas for left/right hands. For body keypoints, we directly use the sigmas provided by the MSCOCO team.

nomaad42 commented 1 year ago

The sigmas are calculated individually/separately for each keypoint. So it is possible that some symmetric parts have different sigmas. For the face keypoints, the sigmas are not symmetric either. For hand keypoints, we treat the left and the right hands equally (for model design, and evaluation). So we specially average the sigmas for left/right hands. For body keypoints, we directly use the sigmas provided by the MSCOCO team.

How do you calculate the sigmas for the each keypoint?