search

Walter0807 / MotionBERT

[ICCV 2023] PyTorch Implementation of "MotionBERT: A Unified Perspective on Learning Human Motion Representations"
Apache License 2.0
1.02k stars 124 forks source link

Smpl Joint rotations. #34

Closed larsh0103 closed 1 year ago

larsh0103 commented 1 year ago

Hey! Thank you for this amazing work! I was wondering if there is any way to extract / estimate the smpl joint rotations when running wild inference? I understand that the refined model actually regresses a joint rotation residual, but I don't see model weights or specific inference code for that one.

Walter0807 commented 1 year ago

Thanks for your interest! You can get the rotations from output[0]['theta']. Code and model weight for the refiner module is in the branch 'refine'.