Image Clustering on CIFAR100 (val)
0.874NMIProbing CLIP ViT-L/14
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Probing CLIP ViT-L/14Backbone=ViT-L/14, Pre-training Strategy=CLIP, Evaluation Protocol=Linear Probing2023.03 | 0.874 | 0.871 | 0.765 | |
| Probing Sup. ViT-L/16Backbone=ViT-L/16, Pre-training Strategy=Supervised, Evaluation Protocol=Linear Probing2023.03 | 0.86 | 0.863 | 0.75 | |
| Probing DINO ViT-B/16Backbone=ViT-B/16, Pre-training Strategy=DINO, Evaluation Protocol=Linear Probing2023.03 | 0.857 | 0.853 | 0.736 | |
| TEMI Sup. ViT-L/16Backbone=ViT-L/16, Pre-training Strategy=Supervised, Evaluation Protocol=Clustering2023.03 | 0.852 | 0.818 | 0.706 | |
| Probing MSN ViT-L/16Backbone=ViT-L/16, Pre-training Strategy=MSN, Evaluation Protocol=Linear Probing2023.03 | 0.846 | 0.844 | 0.719 | |
| TEMI CLIP ViT-L/14Backbone=ViT-L/14, Pre-training Strategy=CLIP (Natural Language Supervised), Evaluation Protocol=Clustering2023.03 | 0.799 | 0.737 | 0.612 | |
| TEMI DINO ViT-B/16Backbone=ViT-B/16, Pre-training Strategy=DINO, Evaluation Protocol=Clustering2023.03 | 0.769 | 0.671 | 0.533 | |
| TEMI MSN ViT-L/16Backbone=ViT-L/16, Pre-training Strategy=MSN, Evaluation Protocol=Clustering2023.03 | 0.73 | 0.614 | 0.474 |