Composed Image Retrieval on CIRR v1.0
70.9Recall@1MagicLens-L
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MagicLens-LParameters=613M, Backbone=CoCa-L2024.03 | 70.9 | 87.3 | 94.5 | |
| MagicLens-BParameters=267M, Backbone=CoCa-B2024.03 | 69.3 | 86 | 94 | |
| Pic2WordParameters=2.5B, Backbone=CLIP-G2024.03 | 68.9 | 85.5 | 93 | |
| SEARLEParameters=2.6B, Backbone=CLIP-G2024.03 | 68.7 | 84.7 | 93.2 | |
| MagicLens-LParameters=465M, Backbone=CLIP-L2024.03 | 68.1 | 84.8 | 93.2 | |
| CIREVLParameters=14.6B*, Backbone=CLIP-G2024.03 | 68 | 84.9 | 93.2 | |
| MagicLens-BParameters=166M, Backbone=CLIP-B2024.03 | 66.7 | 83.9 | 92.4 | |
| SEARLEParameters=1.0B, Backbone=CLIP-H2024.03 | 64.6 | 83.2 | 92.8 | |
| CompoDiffParameters=2.9B, Backbone=CLIP-G2024.03 | 64.5 | 82.4 | 91.8 | |
| LinCIRParameters=2.6B, Backbone=CLIP-G2024.03 | 63.4 | 82.2 | 92 | |
| LinCIRParameters=1.0B, Backbone=CLIP-H2024.03 | 62.4 | 81.5 | 92.1 | |
| Pic2WordParameters=987M, Backbone=CLIP-H2024.03 | 62.2 | 81.4 | 91.2 | |
| CIREVLParameters=12.3B*, Backbone=CLIP-B2024.03 | 60.2 | 80.1 | 90.2 | |
| CIREVLParameters=12.5B*, Backbone=CLIP-L2024.03 | 59.5 | 79.9 | 89.7 | |
| CompoDiffParameters=568M, Backbone=CLIP-L2024.03 | 57.4 | 77.1 | 87.9 | |
| LinCIRParameters=442M, Backbone=CLIP-L2024.03 | 57.1 | 77.4 | 88.9 | |
| PLIParameters=428M†, Backbone=CLIP-L2024.03 | 55.6 | 77.5 | 89.5 | |
| PLIParameters=224M†, Backbone=BLIP-B2024.03 | 55.1 | 77.4 | 89.1 | |
| SEARLEParameters=165M, Backbone=CLIP-B2024.03 | 54.9 | 76.6 | 88.2 | |
| SEARLEParameters=442M, Backbone=CLIP-L2024.03 | 53.8 | 75 | 88.2 | |
| PALAVRAParameters=176M, Backbone=CLIP-B2024.03 | 41.6 | 65.3 | 80.9 |