3D semantic occupancy prediction on nuCraft high-resolution
13.4Overall mIoUC-CONet
Evaluation results
| Method | Links | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| C-CONet3D GT=Dense, Pretraining=-2025.03 | 13.4 | 14.3 | 9.1 | 16.5 | 18.3 | 7.4 | 12.3 | 11.1 | 9.4 | 5.8 | 13.2 | 32.5 | — | — | — | 19.9 | |
| TT-OccLiDAR3D GT=✗, Pretraining=✗2025.03 | 10.92 | 14.92 | 8.15 | 12.21 | 11.93 | 10.71 | 13.82 | 6.25 | 8.39 | 6.93 | 9.05 | 12.57 | 8.41 | 10.29 | 12.89 | 17.37 | |
| TT-OccCamera3D GT=✗, Pretraining=✗2025.03 | 5.95 | 9.21 | 5.58 | 5.35 | 6.19 | 3.93 | 7.28 | 4.86 | 4.73 | 2.89 | 3.44 | 10.15 | 6.17 | 7.18 | 3.92 | 8.45 | |
| SelfOcc3D GT=✗, Pretraining=384 hrs2025.03 | 2.22 | 0.41 | 0.54 | 2.79 | 7.12 | 0 | 0.81 | 1.67 | 0 | 0 | 5.5 | 2.41 | 3.88 | 3.55 | 1.96 | 2.72 |