You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Many thanks for your great work!
I've run the demo scripts successfully and generate the video, however I found that the human hand pose is not predicted correctly, it seems that the pose of hand is initial all the time. And I also notice that you just predict 25 body keypoints by openpose without hand keypoints. I wonder if it's caused by Frankmocap or you just not fit the hand pose.
Thanks again and looking forward to your reply!
The text was updated successfully, but these errors were encountered:
Hi, sorry for the late reply.
Yes, we do not optimize the hand pose here. It is because the hand is quite heavily occluded in lots of frames. So tracking finger poses is quite difficult. If you don't do it correctly then the fingers will look very unatural. Also the BEHAVE images were captured with hands wearing gloves, which makes the 2D/3D hand pose detection quite noise. These problems make the tracking very difficult.
Many thanks for your great work!
![image](https://private-user-images.githubusercontent.com/145037668/278316271-c5d66ce0-d284-4936-b055-85a7321ab493.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3Mzk2MDg1NDIsIm5iZiI6MTczOTYwODI0MiwicGF0aCI6Ii8xNDUwMzc2NjgvMjc4MzE2MjcxLWM1ZDY2Y2UwLWQyODQtNDkzNi1iMDU1LTg1YTczMjFhYjQ5My5wbmc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjUwMjE1JTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI1MDIxNVQwODMwNDJaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT04ZWEyZDAzNjNiN2UzMzJhMDZjMjVhODg3ZDJiNTRhZGE4MzdjM2U5YmM3YTdjZDhiMGIwOGZjZGQ1ZTQ0NWViJlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCJ9.SRza_iek3wAVZof8PgJn7_v-_b5DjtNCdHpdq9zIn_U)
I've run the demo scripts successfully and generate the video, however I found that the human hand pose is not predicted correctly, it seems that the pose of hand is initial all the time. And I also notice that you just predict 25 body keypoints by openpose without hand keypoints. I wonder if it's caused by Frankmocap or you just not fit the hand pose.
Thanks again and looking forward to your reply!
The text was updated successfully, but these errors were encountered: