LiDAR Semantic Segmentation on SemanticSTF (val) (Adverse Weather Metrics)
53.2mIoU (D-fog)OmniLiDAR
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| OmniLiDARBackbone=MinkUNet2026.05 | 53.2 | 56.6 | 59 | 55.3 | 57.9 | |
| sup-onlyBackbone=MinkUNet2026.05 | 52.5 | 55.1 | 58.6 | 54 | 56.2 | |
| OracleTraining source=SynLiDAR2024.07 | 51.9 | 54.6 | 57.9 | 53.7 | 54.7 | |
| PointDR*‡Training source=SynLiDAR, Status=reproduced, Validation set for model selection=SynLiDAR sequence 02024.07 | 21.41 | 20.94 | 25.48 | 19.31 | 20.47 | |
| PointDR*Training source=SynLiDAR, Status=reproduced2024.07 | 19.09 | 20.28 | 25.29 | 18.98 | 19.78 | |
| Baseline SJ+LPDTraining source=SynLiDAR, Augmentation=Selective Jittering + Learnable Point Drop2024.07 | 19.08 | 20.65 | 21.97 | 17.27 | 20.08 | |
| Baseline+SJ+LPD‡Training source=SynLiDAR, Augmentation=Selective Jittering + Learnable Point Drop, Validation set for model selection=SynLiDAR sequence 02024.07 | 18.99 | 21.22 | 23.14 | 17.28 | 20.51 | |
| PolarMixTraining source=SynLiDAR2024.07 | 16.47 | 18.69 | 19.63 | 15.98 | 18.09 | |
| LaserMixTraining source=SynLiDAR2024.07 | 15.32 | 17.95 | 18.55 | 13.8 | 16.85 | |
| BaselineTraining source=SynLiDAR2024.07 | 15.24 | 15.97 | 16.83 | 12.76 | 15.45 |