Open-Vocabulary Semantic Segmentation on Pascal VOC 2012 (val)
96.8mIoUSegEarth-OV3
Evaluation Results
| Method | Links | |
|---|---|---|
| SegEarth-OV3Size=PE-L+/142025.12 | 96.8 | |
| CAT-SegSize=ViT-B/16, Training Protocol=Training-based2025.12 | 94.6 | |
| CorrCLIPSize=ViT-H/14, Training Protocol=Training-free2025.12 | 91.8 | |
| CorrCLIPSize=ViT-L/14, Training Protocol=Training-free2025.12 | 91.5 | |
| CaRSize=ViT-L/14, Training Protocol=Training-free2025.12 | 91.4 | |
| CLIPerSize=ViT-L/14, Training Protocol=Training-free2025.12 | 90 | |
| CorrCLIPSize=ViT-B/16, Training Protocol=Training-free2025.12 | 88.8 | |
| TridentSize=ViT-H/14, Training Protocol=Training-free2025.12 | 88.7 | |
| SC-CLIPSize=ViT-L/14, Training Protocol=Training-free2025.12 | 88.3 | |
| FreeDASize=ViT-L/14, Training Protocol=Training-free2025.12 | 87.9 | |
| CASSSize=ViT-B/16, Training Protocol=Training-free2025.12 | 87.8 | |
| ResCLIPSize=ViT-B/16, Training Protocol=Training-free2025.12 | 86 | |
| ResCLIPSize=ViT-L/14, Training Protocol=Training-free2025.12 | 85.5 | |
| CLIPerSize=ViT-B/16, Training Protocol=Training-free2025.12 | 85.2 | |
| TridentSize=ViT-B/16, Training Protocol=Training-free2025.12 | 84.5 | |
| SC-CLIPSize=ViT-B/16, Training Protocol=Training-free2025.12 | 84.3 | |
| ProxyCLIPSize=ViT-H/14, Training Protocol=Training-free2025.12 | 83.3 | |
| TCLSize=ViT-B/16, Training Protocol=Training-based2025.12 | 83.2 | |
| ProxyCLIPSize=ViT-L/14, Training Protocol=Training-free2025.12 | 83.2 | |
| NACLIPSize=ViT-B/16, Training Protocol=Training-free2025.12 | 83 | |
| LaVGSize=ViT-B/16, Training Protocol=Training-free2025.12 | 82.5 | |
| CLIPtraseSize=ViT-B/16, Training Protocol=Training-free2025.12 | 81.2 | |
| CLIP-DINOiserSize=ViT-B/16, Training Protocol=Training-based2025.12 | 80.9 | |
| ClearCLIPSize=ViT-B/16, Training Protocol=Training-free2025.12 | 80.9 | |
| SCLIPSize=ViT-B/16, Training Protocol=Training-free2025.12 | 80.4 | |
| ProxyCLIPSize=ViT-B/16, Training Protocol=Training-free2025.12 | 80.3 | |
| MaskCLIPSize=ViT-B/16, Training Protocol=Training-free2025.12 | 74.9 | |
| CLIPSize=ViT-B/16, Training Protocol=Training-free2025.12 | 41.9 |