About extra 2D pose datasets (MPII&LSP) data pre-processing and training problem

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 extra 2D pose datasets (MPII&LSP) data pre-processing and training problem #50

Closed wenwen12321 closed 2 years ago

wenwen12321 commented 2 years ago

Hi author. Thanks for the great work!

I'm still not clear on data pre-processing. Could you please release the code or point me how to use additional in-the-wild 2D poses (MPII & LSP) to train the 2D discriminator? Which you mention in paper page 6 "Results on 3DHP (cross-scenario) part".

Garfield-kh commented 2 years ago

Hi, thank you for the interest! Since the 2D dataset are COCO format, you will need to convert it to 16 joints setting. Here are the issue related to the conversion #4 #12 . Meanwhile, we remove some of the 2D data which are too far/small or too close/big to the camera.

wenwen12321 commented 2 years ago

I've got it! Thanks for the very informative response!