search

microsoft / multiview-human-pose-estimation-pytorch

This is an official Pytorch implementation of "Cross View Fusion for 3D Human Pose Estimation, ICCV 2019".
MIT License
549 stars 89 forks source link

In the wild dataset #13

Closed ogata2606 closed 5 years ago

ogata2606 commented 5 years ago

Thank you for opening the source code! I want to try 3d pose estimation in the wild video.

Is it possible to try in the wild dataset?

CHUNYUWANG commented 5 years ago

For 2D pose estimation part, please refer to section 3.2 in the paper for possible workaround. For 3D pose estimation part, it can be directly applied to wild dataset as long as the camera parameters are known.

ogata2606 commented 5 years ago

Thank you reply to my question.

In the paper, it is compared to multiview 3d pose estimation and single view 3d pose estimation. Is the camera parameter necessary even if we estimate 3d pose estimation from a single view?

CHUNYUWANG commented 5 years ago

Yes, the camera parameters are needed even if you use only one camera.

ogata2606 commented 5 years ago

OK. Thank you for responding to the question.

I'll close this issue.

lisa676 commented 5 years ago

@ogata2606 Hi, did you run this code? and what's the result (2D and 3D) if you have run.

ogata2606 commented 5 years ago

@lan786 Yes. I run this code. The H36m data set is fine. But, it does not good in my dataset. I attach 2d pose estimation result with h36m. forward