Text-to-Image Retrieval on ShareGPT4V 5,000 samples (test)
85.48R@1CLIP+GOAL
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CLIP+GOALBackbone=ViT-L/142025.03 | 85.48 | 96.84 | 99.66 | 99.86 | |
| CLIP+GOALBackbone=ViT-B/162025.03 | 79.5 | 94.82 | 99.34 | 99.74 | |
| CLIP+LongCLIPBackbone=ViT-B/162025.03 | 66.86 | 88.72 | 97.56 | 99.2 | |
| CLIP+LongCLIPBackbone=ViT-L/142025.03 | 66.85 | 88.8 | 97.62 | 99.14 | |
| CLIPBackbone=ViT-B/162025.03 | 61.12 | 83.82 | 95.84 | 98.42 | |
| CLIPBackbone=ViT-L/142025.03 | 53.72 | 76.4 | 91.28 | 95.6 |