Text-to-Image Retrieval on MS-COCO (val)
45.1R@1DINOv2-ARL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DINOv2-ARLTraining Samples=20M2024.09 | 45.1 | — | — | |
| LAION-CLIP ViT-LBackbone=ViT-L, Training Samples=400M2024.09 | 43 | — | — | |
| DINOv2-MpNetBackbone=MpNet, Training Samples=20M2024.09 | 42.6 | — | — | |
| OpenAI-CLIP ViT-LBackbone=ViT-L, Training Samples=400M2024.09 | 36.5 | — | — | |
| LilTLwA-baseBackbone=base, Training Samples=0.5M2024.09 | 33.7 | — | — | |
| LiT L16LBackbone=L16L, Training Samples=112M2024.09 | 31.2 | — | — | |
| LilTDA-baseBackbone=base, Training Samples=0.5M2024.09 | 29.1 | — | — | |
| CLIPw/ boxes=false2025.10 | — | 56.29 | 67.53 | |
| CLIPw/ boxes=true2025.10 | — | 56.2 | 67.5 | |
| HyCoCLIPw/ boxes=true2025.10 | — | 57.11 | 68.32 | |
| MERUw/ boxes=false2025.10 | — | 55.73 | 67.02 | |
| MERUw/ boxes=true2025.10 | — | 55.87 | 67.21 | |
| PHyCLIPw/ boxes=true2025.10 | — | 58.03 | 69.05 |