How to Train a Skeleton-based Spatio-Temporal Action Detection Model ?

[Spawnfile]

Hello,

I've generated a skeleton dataset based on the script called ntu_pose_extraction.py to train a PoseC3D and STGCN. However the result I get with demo_skeleton.py is action recognition instead of a spatial temporal action detection. How can I train & inference spatio temporal action detection based on skeleton dataset ?

Cheers