search

Gengzigang / PCT

This is an official implementation of our CVPR 2023 paper "Human Pose as Compositional Tokens" (https://arxiv.org/pdf/2303.11638.pdf)
MIT License
311 stars 20 forks source link

All keypoints sharing the same confidence in pose_results #18

Open zhengthomastang opened 1 year ago

zhengthomastang commented 1 year ago

When I ran demo_img_with_mmdet.py, I checked the output pose_results from inference_top_down_pose_model(). All the 17 keypoints share the same confidence value, which seems to be the aggregated confidence of all 17 keypoints. Could you fix the issue to show individual confidence of each keypoint in the demo script?

Gengzigang commented 1 year ago

Hello, thank you for your interest in our work. Our method treats the pose as a whole, so the predicted score is for the entire pose, and we do not assign individual scores to each keypoints.

morzh commented 11 months ago

Hello, thank you for your interest in our work. Our method treats the pose as a whole, so the predicted score is for the entire pose, and we do not assign individual scores to each keypoints.

Hi ! And what about confidence value itself. Is it in percents? If yes, it always less than 10 even for well illuminated (full body) person without occlusions with near perfect tracked joints.