search

jfzhang95 / PoseAug

[CVPR 2021] PoseAug: A Differentiable Pose Augmentation Framework for 3D Human Pose Estimation, (Oral, Best Paper Award Finalist)
MIT License
368 stars 57 forks source link

About Visualization #54

Closed davidpengiupui closed 1 year ago

davidpengiupui commented 1 year ago

Thanks so much for your wonderful work and help! I used the code in PoseTriplet to do inference. It works very well with the model provided by PoseTriplet. However, when I tried to use the model trained here to test, it doesn't work as their weights are not matching. The problem lies in Line 36 of function_baseline/model_pos_preparation.py, where the output joint number is 15, while that in PoseTriplet is 16. Could you please tell me how to solve it?

Garfield-kh commented 1 year ago

Hi thank you for your interest. The output numbder 15 means it assume the root is 0. To solve it, you need to change the model file in the visualization to follow the same setting as it's training framework.

Hope this helps.

davidpengiupui commented 1 year ago

Thank you for your explanation!