Traversability Estimation on Cityscapes
85.1mIoUViTA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ViTAType=VFM (vision foundation model-based)2026.05 | 85.1 | 85.4 | 95.5 | |
| GeNIEType=VFM (vision foundation model-based)2026.05 | 81.1 | 79 | 98.7 | |
| SAM2 FTType=VFM (vision foundation model-based), Evaluation Protocol=FT (full fine-tuning)2026.05 | 80.7 | 81.2 | 99.2 | |
| SAM2 FZType=VFM (vision foundation model-based), Evaluation Protocol=FZ (encoder frozen)2026.05 | 80.1 | 80.8 | 98.9 | |
| ST-SegType=SS (semantic segmentation)2026.05 | 79.7 | 80.7 | 98.5 | |
| SegFormerType=SS (semantic segmentation)2026.05 | 78.9 | 80.8 | 97.2 | |
| Mask2FormerType=SS (semantic segmentation)2026.05 | 78.7 | 79.8 | 98.2 | |
| DINOv3 FTType=VFM (vision foundation model-based), Evaluation Protocol=FT (full fine-tuning)2026.05 | 76.6 | 78.3 | 97.3 | |
| GANavType=SS (semantic segmentation)2026.05 | 71.6 | 76.7 | 91.6 | |
| STEPPType=SELF (self-supervised traversability)2026.05 | 59.2 | 83.9 | 66.8 | |
| DINOv3 FZType=VFM (vision foundation model-based), Evaluation Protocol=FZ (encoder frozen)2026.05 | 59.1 | 82.6 | 67.5 |