Text-to-Image Retrieval on COCO 2014 (test)
56.78AccuracyNegationCLIP
Evaluation Results
| Method | Links | |
|---|---|---|
| NegationCLIPArch.=ViT-L/142026.03 | 56.78 | |
| NegationCLIPArch.=ViT-B/162026.03 | 53.86 | |
| Omni-NegCLIPArch.=ViT-L/142026.03 | 53.84 | |
| NegationCLIPArch.=ViT-B/322026.03 | 52.12 | |
| Omni-NegCLIPArch.=ViT-B/162026.03 | 51.06 | |
| Omni-NegCLIPArch.=ViT-B/322026.03 | 48.9 | |
| CoN-CLIPArch.=ViT-L/142026.03 | 47.9 | |
| CLIPArch.=ViT-L/142026.03 | 47.61 | |
| CoN-CLIPArch.=ViT-B/162026.03 | 45.38 | |
| CLIPArch.=ViT-B/162026.03 | 42.9 | |
| CoN-CLIPArch.=ViT-B/322026.03 | 41.72 | |
| CLIPArch.=ViT-B/322026.03 | 40.88 |