Zero-Shot Composed Image Retrieval on GeneCIS
18.4Focus Recall@1FoCo
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| FoCoBackbone=CLIP-ViT-G/142026.07 | 18.4 | 27.9 | 39 | 19.3 | 31.8 | 40 | 19.6 | 32.3 | 41 | 18.1 | 27.4 | 38 | |
| FoCoBackbone=CLIP-ViT-L/142026.07 | 17.9 | 26.9 | 37.2 | 17.9 | 28.1 | 35.7 | 19.3 | 31.5 | 39.7 | 16.6 | 23.5 | 33.2 | |
| PrediCIRBackbone=CLIP-ViT-G/142026.07 | 16.1 | 26.3 | 37.5 | 19.4 | 31.5 | 39.8 | 19.6 | 32 | 40.8 | 15.9 | 25.8 | 36.5 | |
| PrediCIRBackbone=CLIP-ViT-L/142026.07 | 15.5 | 25.5 | 36.9 | 17.8 | 28 | 34.8 | 18.5 | 31.2 | 39 | 14.8 | 22.3 | 32.7 | |
| LinCIRBackbone=CLIP-ViT-G/142026.07 | 14.6 | 26.1 | 35.2 | 12.8 | 23.3 | 31.9 | 18.4 | 31.6 | 40.2 | 9 | 17.7 | 26.9 | |
| Context-I2WBackbone=CLIP-ViT-L/142026.07 | 13 | 24.2 | 34.3 | 12.1 | 22.2 | 31.3 | 16.8 | 29.4 | 39.4 | 8.2 | 17 | 26.2 | |
| LinCIRBackbone=CLIP-ViT-L/142026.07 | 12.6 | 23.7 | 33.9 | 11.8 | 21.9 | 30.9 | 16.6 | 29 | 39.2 | 7.9 | 16.6 | 25.6 | |
| SEARLE-XLBackbone=CLIP-ViT-L/142026.07 | 12.4 | 23.2 | 33 | 12.2 | 21.1 | 29.6 | 16.7 | 27.5 | 37.4 | 7.9 | 16.8 | 25.2 | |
| Pic2WordBackbone=CLIP-ViT-L/142026.07 | 12.1 | 23.1 | 32.3 | 10.3 | 19.9 | 28.6 | 14.8 | 26.5 | 35.9 | 7.6 | 16.6 | 24.9 |