Clustering on ImageNet (test)
72.4AccuracyTURTLE
Evaluation results
| Method | Links | |
|---|---|---|
| TURTLEBackbone=Concatenated CLIP ViT-L/14 + DINOv2 ViT-g/142024.06 | 72.4 | |
| HUME DINOv2Backbone=DINOv2 ViT-g/142024.06 | 68.1 | |
| HUME CLIPBackbone=CLIP ViT-L/142024.06 | 65.2 | |
| K-MeansBackbone=Concatenated CLIP ViT-L/14 + DINOv2 ViT-g/142024.06 | 64.8 |