search

facebookresearch / VideoPose3D

Efficient 3D human pose estimation in video using 2D keypoint trajectories
Other
3.69k stars 748 forks source link

How to convert predicted points from VideoPose3d to use on the image frame #125

Open 97k opened 4 years ago

97k commented 4 years ago

Hi! I loved this project and its a great work. Thank you for open-sourcing it!

I've been using this project to find joint co-ordinates and as far I've got the following

I am trying to convert the coordinates (which are in meters right?) into real world(image space) coordinates in the wild videos.

The skeleton reconstruction is perfect but I am not able to convert the project the 3d co-ordinates into 2D (similar to output of keypoint RCNN). In other words, I am trying to use the x,y coordinates of the joints predicted from VideoPose3D instead of KeypointsRCNN.

Can you please provide a little bit more information on how that can be achieved?

Pardon me for the inconvenience, Should I provide further details, please let me know.

Thank You!

kallivad commented 4 years ago

Have the same question! :)

eylonmiz22 commented 3 years ago

Me too, the same problem. Did someone already solve this?