Arthur151 / ROMP

Monocular, One-stage, Regression of Multiple 3D People and their 3D positions & trajectories in camera & global coordinates. ROMP[ICCV21], BEV[CVPR22], TRACE[CVPR2023]
https://www.yusun.work/
Apache License 2.0
1.36k stars 231 forks source link

The meaning of function "convert_trans2cam" #507

Open coding-rachal opened 7 months ago

coding-rachal commented 7 months ago

Hi, sorry to bother you. I want to train the model based on a new dataset captured by moving cameras, Which coordinates our initial 3dkpts should under for the data loader to load?