How to use the BEV representation?

huang-yh / SelfOcc

[CVPR 2024] SelfOcc: Self-Supervised Vision-Based 3D Occupancy Prediction

Apache License 2.0

273 stars 17 forks source link

How to use the BEV representation? #10

Open rmarcuzzi opened 7 months ago

rmarcuzzi commented 7 months ago

Hi!

Thanks a lot for your work! I was wondering if you could explain where to modify the code to use the BEV representation instead of the TPV. In the config, I can see that the TPV segmentor is chosen here and also in many other configurations I can see you're using TPV (although I see some commented lines using BEV). Also in model/segmentor I can only see the TPV segmentor.

Thanks a lot!

huang-yh commented 3 months ago

I just uploaded an example config using bev for occupancy prediction on nuScenes (config/nuscenes/nuscenes_occ_bev.py). Reasonable performance is expected, but I am not sure whether it could achieve the result we reported in the paper.