Unsupervised Learning on ImageNet-1K
43.4AccuracyiBOT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| iBOTArch.=ViT-S/16, Protocol=k-means clustering on frozen features2021.11 | 43.4 | 32.8 | 78.6 | 35.6 | |
| DINOArch.=ViT-S/16, Protocol=k-means clustering on frozen features2021.11 | 41.4 | 29.8 | 76.8 | 32.8 | |
| SCANArch.=RN50, Protocol=k-means clustering on frozen features2021.11 | 39.9 | 27.5 | 72 | — | |
| InfoMinArch.=RN50, Protocol=k-means clustering on frozen features2021.11 | 33.2 | 14.7 | 68.8 | — | |
| Self-labelArch.=RN50, Protocol=k-means clustering on frozen features2021.11 | 30.5 | 16.2 | 75.4 | — |