search

Vegetebird / MHFormer

[CVPR 2022] MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation
MIT License
530 stars 85 forks source link

Something about vis.py #90

Closed Ared521 closed 1 year ago

Ared521 commented 1 year ago

Hi author, I'm running vis.py, and I want to get three-dimensional coordinate information x,y,z. I see there are post_out and output_3D in the code, may I ask which coordinate is it? What's the difference between these two parameters?

Vegetebird commented 1 year ago

The output_3D is in camera coordinate, and post_out is in world coordinate. Maybe post_out is more appropriate for you.

Ared521 commented 1 year ago

The output_3D is in camera coordinate, and post_out is in world coordinate. Maybe post_out is more appropriate for you.

OK, Thank you very much.

jerrinbright commented 1 year ago

Hey! I have a query in regard to the coordinate system. Did you use World or Camera coordinate pose data for training? Also, will using the pose data in different coordinate systems affect the model's performance?

mnauf commented 1 year ago

@jerriebright The author has used a camera coordinate system for training. The models predict 3d joint locations in the camera coordinate system, and later the author explicitly converts them to a world coordinate system. This step is optional

jerrinbright commented 1 year ago

Hi @mnauf , thank you!