search

Daniil-Osokin / lightweight-human-pose-estimation-3d-demo.pytorch

Real-time 3D multi-person pose estimation demo in PyTorch. OpenVINO backend can be used for fast inference on CPU.
Apache License 2.0
653 stars 137 forks source link

the trainset #75

Closed ZXin0305 closed 3 years ago

ZXin0305 commented 3 years ago

hello,thanks for your great work. I want to know that how to build the trainset about “ORPM”? I want to train it with another method while I don't know how to do this. Thank you very much!

Daniil-Osokin commented 3 years ago

Hi, check the #17 for details about ORPM. We have used CMU Panoptic dataset to train 3D pose estimation.

Daniil-Osokin commented 3 years ago

Hope, it helped.

nicolasugrinovic commented 2 years ago

Hi @Daniil-Osokin ,

Kudos for the great work here to make this model real-tme. Just to make it very clear, you only train on Panoptic and COCO right? You don't train in MuCO-3DHP as in the "Single-Shot Multi-Person 3D Pose Estimation From Monocular RGB " paper?

Thanks,

Daniil-Osokin commented 2 years ago

Hi! Yes, the model was not trained on MuCO-3DHP (it was hard to obtain it in that time).