Semantic Segmentation on nuImages
84.2mIoUDINOv3 + CD
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DINOv3 + CDArch.=ViT-H+/16, Evaluation Protocol=linear probing2026.02 | 84.2 | — | |
| DINOv3Arch.=ViT-H+/16, Evaluation Protocol=linear probing2026.02 | 83.7 | — | |
| DepthAnythingV2 + CDArch.=ViT-B/14, Evaluation Protocol=linear probing2026.02 | 75.8 | — | |
| DepthAnythingV2Arch.=ViT-B/14, Evaluation Protocol=linear probing2026.02 | 75.1 | — | |
| CLIP + CDArch.=ViT-B/16, Evaluation Protocol=linear probing2026.02 | 67.9 | — | |
| IBOT + CDArch.=ViT-L/16, Evaluation Protocol=linear probing2026.02 | 67.7 | — | |
| DINO + CDArch.=ViT-B/8, Evaluation Protocol=linear probing2026.02 | 66.1 | — | |
| IBOTArch.=ViT-L/16, Evaluation Protocol=linear probing2026.02 | 65.9 | — | |
| CLIPArch.=ViT-B/16, Evaluation Protocol=linear probing2026.02 | 65.8 | — | |
| DINOArch.=ViT-B/8, Evaluation Protocol=linear probing2026.02 | 62.4 | — | |
| DINO + CDArch.=ResNet50, Evaluation Protocol=linear probing2026.02 | 61.3 | — | |
| VE-Promptmulti-task=true2023.03 | 59.1 | 57.5 | |
| DINO basedmulti-task=true2023.03 | 56.7 | 56.1 | |
| MoCov2 + CDArch.=ResNet50, Evaluation Protocol=linear probing2026.02 | 55.8 | — | |
| DINOArch.=ResNet50, Evaluation Protocol=linear probing2026.02 | 55.8 | — | |
| MoCov2Arch.=ResNet50, Evaluation Protocol=linear probing2026.02 | 55.3 | — | |
| Sparse R-CNN basedmulti-task=true2023.03 | 53.8 | 52.1 |