Bounding box classification on COCO 80
77.4Top-1 AccuracyFG-CLIP 2
Evaluation Results
| Method | Links | |
|---|---|---|
| FG-CLIP 2Backbone=ViT-So/162025.10 | 77.4 | |
| FG-CLIP 2Backbone=ViT-B/162025.10 | 74.9 | |
| FG-CLIP 2Backbone=ViT-L/162025.10 | 74 | |
| CLOCBackbone=ViT-L/142025.10 | 72.9 | |
| FG-CLIPBackbone=ViT-L/142025.10 | 63.2 | |
| SigLIP 2Backbone=ViT-So/162025.10 | 62 | |
| SigLIP 2Backbone=ViT-L/162025.10 | 54.7 | |
| FineCLIPBackbone=ViT-L/142025.10 | 54.5 | |
| SigLIP 2Backbone=ViT-B/162025.10 | 53.4 | |
| FG-CLIPBackbone=ViT-B/162025.10 | 52.3 | |
| Meta CLIP 2Backbone=ViT-H/142025.10 | 52 | |
| FineCLIPBackbone=ViT-B/162025.10 | 48.4 | |
| CLIPBackbone=ViT-B/162025.10 | 44.2 | |
| Long-CLIPBackbone=ViT-B/162025.10 | 36.7 | |
| Long-CLIPBackbone=ViT-L/142025.10 | 35.6 | |
| CLIPBackbone=ViT-L/142025.10 | 33.8 | |
| EVA-CLIPBackbone=ViT-L/142025.10 | 32.1 | |
| EVA-CLIPBackbone=ViT-B/162025.10 | 30.6 |