Semantic occupancy prediction on Occ3D-nuScenes
23.5IoU (Semantic)QueryOcc+
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| QueryOcc+backbone=ConvNeXt-Base, Input Resolution=high-res, Supervision source=camera and lidar, Feature distillation=VFM2025.11 | 23.5 | 25.8 | 23.8 | 47.4 | 15.7 | 56.9 | |
| QueryOccbackbone=ConvNeXt-Base2025.11 | 21.3 | 23.6 | 21.7 | 45.2 | 13.2 | 55 | |
| GaussianFlowOcc2025.11 | 17.1 | 18.7 | — | — | 10.1 | 46.9 | |
| GaussianFlowOccreproduction=RayIoU and IoU2025.11 | 16.1 | 18.2 | 17.2 | 36 | 9.9 | 40.2 | |
| GaussTRvariant=-T2D, reproduction=RayIoU2025.11 | 13.9 | 14.2 | 17.7 | 33.8 | 13.4 | 44.5 | |
| LangOccreproduction=RayIoU2025.11 | 13.3 | 11.6 | 9 | 38.7 | 7.7 | 51.8 | |
| GaussTRvariant=-FeatUp, reproduction=RayIoU2025.11 | 13.3 | 13.8 | 14.5 | 34.2 | 9 | 45.2 | |
| MinkOccsupervision=w/lidar2025.11 | 13.2 | 12.5 | — | — | 3.4 | — | |
| GaussianOcc2025.11 | 11.3 | 11.9 | — | — | 7 | — | |
| OccNeRF2025.11 | 10.8 | — | — | — | 3.7 | 22.8 | |
| SelfOccreproduction=RayIoU2025.11 | 10.5 | 10.9 | 7.2 | 29.2 | 3.7 | 45 | |
| DistillNeRF2025.11 | 10.1 | — | — | — | 5.2 | 29.1 |