Open-vocabulary semantic segmentation on VDD
0.675mIoUProC-SAM3
Evaluation Results
| Method | Links | |
|---|---|---|
| ProC-SAM3Backbone=SAM 32026.06 | 0.675 | |
| SegEarth-OV3*Backbone=SAM 32026.06 | 0.644 | |
| OracleEvaluation Protocol=Fully supervised2026.02 | 0.629 | |
| ActiveSAM*Backbone=SAM32026.06 | 0.622 | |
| VIP*Backbone=DINOv3 + dino.txt2026.06 | 0.525 | |
| ConInferBackbone=CLIP+DINOv32026.06 | 0.503 | |
| ReAttnCLIPBackbone=CLIP2026.06 | 0.497 | |
| ProxyCLIPBackbone=CLIP+DINOv22026.06 | 0.478 | |
| CorrCLIPBackbone=CLIP+DINO+SAM22026.06 | 0.473 | |
| OursEvaluation Protocol=Zero-shot2026.02 | 0.468 | |
| SegEarth-OVEvaluation Protocol=Trained on remote sensing data2026.02 | 0.453 | |
| SegEarth-OVBackbone=CLIP2026.06 | 0.453 | |
| RSCLIP*Backbone=CLIP2026.06 | 0.425 | |
| GEMEvaluation Protocol=Zero-shot2026.02 | 0.395 | |
| ClearCLIPEvaluation Protocol=Zero-shot2026.02 | 0.393 | |
| ClearCLIPBackbone=CLIP2026.06 | 0.393 | |
| SCLIPEvaluation Protocol=Zero-shot2026.02 | 0.379 | |
| SCLIPBackbone=CLIP2026.06 | 0.379 | |
| MaskCLIPEvaluation Protocol=Zero-shot2026.02 | 0.329 | |
| CLIPEvaluation Protocol=Zero-shot2026.02 | 0.142 |