search

huang-yh / SelfOcc

[CVPR 2024] SelfOcc: Self-Supervised Vision-Based 3D Occupancy Prediction
Apache License 2.0
273 stars 17 forks source link

Is this method generalizable? #15

Closed Fanqyu closed 2 months ago

Fanqyu commented 4 months ago

Hi, thanks for your great work. As far as I know, NeuS is optimized per scene and is not generalizable. Thus, I wonder if the NeuS used in your work is adapted to a generalizable form. If yes, could you explain the implementation briefly? It could help me save a considerable amount of time. Looking forward to your reply.

huang-yh commented 3 months ago

Yes, SelfOcc is generalizable to novel scenes. At the core of the implementation is a feed-forward TPV encoder self-supervised by depth reprojection loss and neural rendering loss.

Fanqyu commented 2 months ago

Thanks for your reply!