Re-identification on CelebA
95.6AUCCLIP
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| CLIPMethod Category=Foundation Models, Sensitivity=0.992025.12 | 95.6 | 53.9 | — | — | — | — | — | — | |
| LCMemMethod Category=Task-specific, Sensitivity=0.99, Training Protocol=Jointly trained on full set of datasets2025.12 | 95.5 | 64.5 | — | — | — | — | — | — | |
| LCMem Single StageMethod Category=Task-specific, Sensitivity=0.99, Training Protocol=Jointly trained on full set of datasets2025.12 | 95 | 62.5 | — | — | — | — | — | — | |
| ConvNeXtMethod Category=Foundation Models, Sensitivity=0.992025.12 | 88 | 19.6 | — | — | — | — | — | — | |
| DINOv3Method Category=Foundation Models, Sensitivity=0.992025.12 | 82.6 | 17 | — | — | — | — | — | — | |
| DINOv2Method Category=Foundation Models, Sensitivity=0.992025.12 | 80.6 | 11.7 | — | — | — | — | — | — | |
| SwAVMethod Category=Foundation Models, Sensitivity=0.992025.12 | 76.3 | 6.7 | — | — | — | — | — | — | |
| InceptionMethod Category=Foundation Models, Sensitivity=0.992025.12 | 69.3 | 5.8 | — | — | — | — | — | — | |
| MAEMethod Category=Foundation Models, Sensitivity=0.992025.12 | 61.6 | 2.1 | — | — | — | — | — | — | |
| BYOLMethod Category=Foundation Models, Sensitivity=0.992025.12 | 60.1 | 1.6 | — | — | — | — | — | — | |
| NCCMethod Category=Pixel-wise, Sensitivity=0.992025.12 | 59.8 | 1.8 | — | — | — | — | — | — | |
| SSIMMethod Category=Pixel-wise, Sensitivity=0.992025.12 | 58.1 | 1.6 | — | — | — | — | — | — | |
| MSEMethod Category=Pixel-wise, Sensitivity=0.992025.12 | 57.9 | 1.3 | — | — | — | — | — | — | |
| RandomMethod Category=Unsupervised, Sensitivity=0.992025.12 | 56.8 | 2.6 | — | — | — | — | — | — | |
| Siamese + NT-XentMethod Category=Task-specific, Sensitivity=0.99, Training Protocol=Jointly trained on full set of datasets2025.12 | 56.3 | 1.8 | — | — | — | — | — | — | |
| Siamese + NT-XentMethod Category=Unsupervised, Sensitivity=0.992025.12 | 55.8 | 1.7 | — | — | — | — | — | — | |
| SiameseMethod Category=Task-specific, Sensitivity=0.99, Training Protocol=Jointly trained on full set of datasets2025.12 | 55 | 1.7 | — | — | — | — | — | — | |
| data2vecMethod Category=Foundation Models, Sensitivity=0.992025.12 | 53.7 | 1.3 | — | — | — | — | — | — | |
| Real2018.11 | — | — | 100 | 100 | 100 | 100 | 100 | 100 | |
| StarGANlambda_cls=0.252018.11 | — | — | 65 | 60 | 64 | 66 | 68 | 65 | |
| StarGAN+lambda_cls=0.252018.11 | — | — | 72 | 66 | 67 | 78 | 69 | 70 | |
| WarpGAN+lambda_cls=2.002018.11 | — | — | 83 | 73 | 81 | 87 | 82 | 81 |