-
Notifications
You must be signed in to change notification settings - Fork 90
Description
Hello. I have a question for the Pose2Sim development team.
Currently, when using Pose2Sim's Whole-body model, I understand it is based on the COCO 133 keypoint model. This model uses only two keypoints per hand, totaling four keypoints, in addition to the torso and foot keypoints.
However, Hand Pose Estimation typically tracks 21 keypoints per hand, totaling 42 keypoints for both hands.
I'm curious if there is a specific reason why only two hand keypoints are used in the 3D modeling process, even though all 133 keypoints can be extracted from the 2D image.
Additionally, I would like to ask if it is possible to add 21 keypoints per hand, totaling 42 hand keypoints, to the body + feet keypoints to achieve more detailed motion analysis.
I look forward to your response. Thank you.