Image-to-Text Retrieval on MS-COCO 5K
0.723R@1SoftMask++
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SoftMask++#Img=4M, Zero-shot=true2023.04 | 0.723 | 0.915 | 0.957 | — | — | |
| CODIS#Img=4M, Zero-shot=true2023.04 | 0.715 | 0.911 | 0.955 | — | — | |
| TCL#Img=4M, Zero-shot=true2023.04 | 0.714 | 0.908 | 0.954 | — | — | |
| ALBEF#Img=4M, Zero-shot=true2023.04 | 0.687 | 0.895 | 0.947 | — | — | |
| UNITER#Img=4M, Zero-shot=true2023.04 | 0.641 | 0.877 | 0.933 | — | — | |
| ReConNoise=0%2025.02 | 0.616 | 0.867 | 0.927 | — | 441.6 | |
| CRCL2023.10 | 0.613 | 0.858 | 0.927 | — | — | |
| ReConNoise=20%2025.02 | 0.611 | 0.857 | 0.922 | — | 437.6 | |
| CRCLNoise Rate=20%, Evaluation Protocol=Fine-tune2023.10 | 0.593 | 0.852 | 0.919 | — | — | |
| DECL2023.10 | 0.592 | 0.845 | 0.915 | — | — | |
| BiCroreproduced=true2023.10 | 0.59 | 0.844 | 0.917 | — | — | |
| GSCNoise level=20%, Evaluation protocol=Finetune2024.05 | 0.589 | 0.849 | 0.917 | 430.8 | — | |
| SGRAF2023.10 | 0.588 | 0.848 | 0.921 | — | — | |
| ALIGN#Img=1.2B, Zero-shot=true2023.04 | 0.586 | 0.83 | 0.897 | — | — | |
| CLIP#Img=400M, Zero-shot=true2023.04 | 0.584 | 0.815 | 0.881 | — | — | |
| CLIPNoise Rate=0%, Evaluation Protocol=Zero-Shot, Backbone=ViT-L/142023.10 | 0.584 | 0.815 | 0.881 | — | — | |
| CLIP-LNoise level=0%, Evaluation protocol=Zeroshot, Backbone=ViT-L/142024.05 | 0.584 | 0.815 | 0.881 | 400.4 | — | |
| CLIP-14Noise=0%, Backbone=ViT-L/142025.02 | 0.584 | 0.815 | 0.881 | — | 400.4 | |
| NCR2023.10 | 0.582 | 0.842 | 0.915 | — | — | |
| ReConNoise=50%2025.02 | 0.581 | 0.851 | 0.919 | — | 428.3 | |
| REPAIRNoise Level=20%, Re-implemented=false2024.03 | 0.576 | 0.841 | 0.914 | 425 | — | |
| MSCNNoise Level=20%, Re-implemented=false2024.03 | 0.571 | 0.84 | 0.908 | 423 | — | |
| NCRNoise level=20%, Evaluation protocol=Finetune2024.05 | 0.569 | 0.836 | 0.91 | 422 | — | |
| VSE∞2023.10 | 0.566 | 0.836 | 0.914 | — | — | |
| ViLT#Img=4M, Zero-shot=true2023.04 | 0.565 | 0.826 | 0.896 | — | — | |
| NCRNoise Level=20%, Re-implemented=true2024.03 | 0.563 | 0.834 | 0.908 | 421.4 | — | |
| GSCNoise level=50%, Evaluation protocol=Finetune2024.05 | 0.555 | 0.818 | 0.901 | 416.3 | — | |
| REPAIRNoise Level=40%, Re-implemented=false2024.03 | 0.555 | 0.828 | 0.901 | 415.4 | — | |
| NCRNoise Level=40%, Re-implemented=true2024.03 | 0.551 | 0.82 | 0.9 | 413.1 | — | |
| BiCroNoise Level=20%, Re-implemented=true2024.03 | 0.533 | 0.818 | 0.897 | 410.9 | — | |
| NCRNoise level=50%, Evaluation protocol=Finetune2024.05 | 0.531 | 0.807 | 0.885 | 404.6 | — | |
| REPAIRNoise Level=60%, Re-implemented=false2024.03 | 0.525 | 0.799 | 0.882 | 400.3 | — | |
| CLIP#Fine-grained alignment=false, Zero-shot=true, Backbone=CLIP ViT-B/16, Training Dataset=untrained2025.11 | 0.523 | 0.762 | — | — | — | |
| BiCroNoise Level=40%, Re-implemented=true2024.03 | 0.507 | 0.794 | 0.882 | 398.9 | — | |
| NCRNoise Level=60%, Re-implemented=true2024.03 | 0.504 | 0.785 | 0.879 | 394.9 | — | |
| CLIPNoise Rate=0%, Evaluation Protocol=Zero-Shot, Backbone=ViT-B/322023.10 | 0.502 | 0.746 | 0.836 | — | — | |
| CLIP-BNoise level=0%, Evaluation protocol=Zeroshot, Backbone=ViT-B/322024.05 | 0.502 | 0.746 | 0.836 | 361.6 | — | |
| CLIP-32Noise=0%, Backbone=ViT-B/322025.02 | 0.502 | 0.746 | 0.836 | — | 361.6 | |
| BiCroNoise Level=60%, Re-implemented=true2024.03 | 0.475 | 0.768 | 0.857 | 381.6 | — | |
| SEPSFine-grained alignment=true, Zero-shot=true, Backbone=CLIP ViT-B/16, Training Dataset=Flickr2025.11 | 0.458 | 0.602 | — | — | — | |
| SCAN2023.10 | 0.447 | 0.759 | 0.866 | — | — | |
| CLIPNoise Rate=20%, Evaluation Protocol=Fine-tune, Backbone=ViT-L/142023.10 | 0.361 | 0.613 | 0.725 | — | — | |
| CLIP-14Noise=20%, Backbone=ViT-L/142025.02 | 0.361 | 0.613 | 0.725 | — | 289.4 | |
| LAPSFine-grained alignment=true, Zero-shot=true, Backbone=CLIP ViT-B/16, Training Dataset=Flickr2025.11 | 0.271 | 0.505 | — | — | — | |
| CLIPNoise Rate=20%, Evaluation Protocol=Fine-tune, Backbone=ViT-B/322023.10 | 0.214 | 0.496 | 0.633 | — | — | |
| CLIP-BNoise level=20%, Evaluation protocol=Finetune, Backbone=ViT-B/322024.05 | 0.214 | 0.496 | 0.633 | 236.3 | — | |
| CLIP-32Noise=20%, Backbone=ViT-B/322025.02 | 0.214 | 0.496 | 0.633 | — | 236.3 | |
| D2S-VSEFine-grained alignment=false, Zero-shot=true, Backbone=CLIP ViT-B/16, Training Dataset=Flickr2025.11 | 0.159 | 0.347 | — | — | — | |
| SCANFine-grained alignment=true, Zero-shot=true, Backbone=CLIP ViT-B/16, Training Dataset=Flickr2025.11 | 0.148 | 0.329 | — | — | — | |
| VSE++Fine-grained alignment=false, Zero-shot=true, Backbone=CLIP ViT-B/16, Training Dataset=Flickr2025.11 | 0.132 | 0.305 | — | — | — | |
| CLIP-BNoise level=50%, Evaluation protocol=Finetune, Backbone=ViT-B/322024.05 | 0.109 | 0.278 | 0.383 | 131.1 | — | |
| CLIP-32Noise=50%, Backbone=ViT-B/322025.02 | 0.109 | 0.278 | 0.383 | — | 131.1 |