open-mmlab / mmpose

OpenMMLab Pose Estimation Toolbox and Benchmark.
https://mmpose.readthedocs.io/en/latest/
Apache License 2.0
5.58k stars 1.21k forks source link

How to reconstruct 3d human body from multiple perspectives? #1735

Open graceforlife opened 1 year ago

graceforlife commented 1 year ago

I want to input videos from multiple perspectives (such as three videos) into the algorithm"3D human pose "to get a 3D result. What should I do?

ly015 commented 1 year ago

Please refer to our multi-view demo. Please note that the camera parameter is required, and the model may not generalize very well on custom data which has different camera parameters from the training data.

graceforlife commented 1 year ago

Please refer to our multi-view demo. Please note that the camera parameter is required, and the model may not generalize very well on custom data which has different camera parameters from the training data.

okok,thanks u~

graceforlife commented 1 year ago

Please refer to our multi-view demo. Please note that the camera parameter is required, and the model may not generalize very well on custom data which has different camera parameters from the training data.

hello ,I have tried what you told,And it works very well,but can the multi-vieo run with the videos instead of images?

ly015 commented 1 year ago

Is it possible to feed video data frame-by-frame into the model? I guess the frames from different cameras can be temporally aligned with timestamps.