Text Retrieval on MSCOCO (5K)
83.2R@1CLIP
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CLIPEvaluation Protocol=Fine-tuned2022.05 | 83.2 | 87.6 | 90 | |
| CYCLIPEvaluation Protocol=Fine-tuned2022.05 | 83.2 | 87.8 | 90.3 | |
| CLIPEvaluation Protocol=Zero-shot2022.05 | 82.1 | 85.6 | 87.8 | |
| CYCLIPEvaluation Protocol=Zero-shot2022.05 | 82.1 | 85.6 | 87.7 | |
| ALIGN#Images=1.2B, Fine-tuning=true2022.02 | 77 | 93.5 | 96.9 | |
| ALIGN#Img=1.2B, fine-tuned=true2023.04 | 77 | 93.5 | 96.9 | |
| SoftMask++#Img=4M, fine-tuned=true2023.04 | 76.6 | 93.5 | 96.6 | |
| Triple Contrastive Learning (TCL)#Images=4M, Fine-tuning=true2022.02 | 75.6 | 92.8 | 96.7 | |
| TCL#Img=4M, fine-tuned=true2023.04 | 75.6 | 92.8 | 96.7 | |
| VinVL#Img=6M, fine-tuned=true2023.04 | 75.4 | 92.9 | 96.2 | |
| CODIS#Img=4M, fine-tuned=true2023.04 | 75.3 | 92.6 | 96.6 | |
| ALBEF#Images=4M, Fine-tuning=true2022.02 | 73.1 | 91.4 | 96 | |
| ALBEF#Img=4M, fine-tuned=true2023.04 | 73.1 | 91.4 | 96 | |
| TCL#Images=4M, Zero-shot=true2022.02 | 71.4 | 90.8 | 95.4 | |
| OSCAR#Images=4M, Fine-tuning=true2022.02 | 70 | 91.1 | 95.5 | |
| OSCAR#Img=4M, fine-tuned=true2023.04 | 70 | 91.1 | 95.5 | |
| ALBEF#Images=4M, Zero-shot=true2022.02 | 68.7 | 89.5 | 94.7 | |
| ImageBERT#Images=6M, Fine-tuning=true2022.02 | 66.4 | 89.8 | 94.4 | |
| SOHO#Images=200K, Fine-tuning=true2022.02 | 66.4 | 88.2 | 93.8 | |
| SOHO#Img=200K, fine-tuned=true2023.04 | 66.4 | 88.2 | 93.8 | |
| UNITER#Images=4M, Fine-tuning=true2022.02 | 65.7 | 88.6 | 93.8 | |
| UNITER#Img=4M, fine-tuned=true2023.04 | 65.7 | 88.6 | 93.8 | |
| UNITER#Images=4M, Zero-shot=true2022.02 | 64.1 | 87.7 | 93.3 | |
| ViLT#Images=4M, Fine-tuning=true2022.02 | 61.5 | 86.3 | 92.7 | |
| ViLT#Img=4M, fine-tuned=true2023.04 | 61.5 | 86.3 | 92.7 | |
| ALIGN#Images=1.2B, Zero-shot=true2022.02 | 58.6 | 83 | 89.7 | |
| CLIP#Images=400M, Zero-shot=true2022.02 | 58.4 | 81.5 | 88.1 | |
| ViLT#Images=4M, Zero-shot=true2022.02 | 56.5 | 82.6 | 89.6 | |
| ImageBERT#Images=6M, Zero-shot=true2022.02 | 44 | 71.2 | 80.4 |