Image Classification on Pets 37
94.8Top-1 AccuracyDinoV2
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DinoV2Architecture=ViT-S/14+reg (22M), Dataset (Size)=LVD (142M), Evaluation Protocol=Frozen features2025.02 | 94.8 | 99.9 | |
| I-JEPAArchitecture=ViT-H/14 (630M), Dataset (Size)=ImageNet-1k (1.3M), Evaluation Protocol=Frozen features2025.02 | 91.7 | 99.2 | |
| SCOTT-12/16Features=MIM-JEPA, Dataset (Size)=†, Evaluation Protocol=Frozen features2025.02 | 90.7 | 99.4 | |
| SCOTT-12/16Architecture=SCOTT-12/16, Params=22 M, Evaluation Protocol=Frozen evaluation on top of MIM-JEPA pretraining, Training Duration=1200 epochs2025.02 | 90.7 | 99.4 | |
| SCOTT-7/16Features=MIM-JEPA, Dataset (Size)=†, Evaluation Protocol=Frozen features2025.02 | 88 | 99 | |
| SCOTT-7/16Architecture=SCOTT-7/16, Params=14 M, Evaluation Protocol=Frozen evaluation on top of MIM-JEPA pretraining, Training Duration=1200 epochs2025.02 | 88 | 99 | |
| SCOTT-12/16Architecture=SCOTT-12/16, Params=22 M, Evaluation Protocol=Frozen evaluation on top of MIM-JEPA pretraining2025.02 | 86.2 | 98.5 | |
| SCOTT-7/16Architecture=SCOTT-7/16, Params=14 M, Evaluation Protocol=Frozen evaluation on top of MIM-JEPA pretraining2025.02 | 81.6 | 97.4 | |
| SCOTT-12/16Architecture=SCOTT-12/16, Params=22 M, Evaluation Protocol=Supervised learning from scratch2025.02 | 67.4 | 90.5 | |
| SCOTT-7/16Architecture=SCOTT-7/16, Params=14 M, Evaluation Protocol=Supervised learning from scratch2025.02 | 67.2 | 89.3 | |
| ViT-12/16Architecture=ViT-12/16, Params=22 M, Evaluation Protocol=Supervised learning from scratch2025.02 | 48.3 | 78.5 |