Image Classification on ImageNet-1K 16-shot (train)
85.9AccuracyADDP
Evaluation results
| Method | Links | |
|---|---|---|
| ADDPBackbone=VIT-L, Evaluation Protocol=Fine-tuning2024.04 | 85.9 | |
| Data2VecBackbone=ViT-B, Evaluation Protocol=Fine-tuning2024.04 | 84.2 | |
| CAEBackbone=ViT-B, Evaluation Protocol=Fine-tuning2024.04 | 83.8 | |
| Con MIMBackbone=ViT-B, Evaluation Protocol=Fine-tuning2024.04 | 83.7 | |
| MAEBackbone=ViT-B, Evaluation Protocol=Fine-tuning2024.04 | 83.6 | |
| DINOBackbone=ViT-B, Evaluation Protocol=Fine-tuning2024.04 | 83.3 | |
| MoCo-v3Backbone=ViT-B, Evaluation Protocol=Fine-tuning2024.04 | 83.2 | |
| Latent MIMBackbone=ViT-B, Evaluation Protocol=Fine-tuning2024.04 | 83 | |
| MAE (αmin = 0.6)Backbone=ViT-B, Evaluation Protocol=Fine-tuning2024.04 | 82.9 | |
| DINOBackbone=ViT-B, Evaluation Protocol=Linear Probing2024.04 | 78.2 | |
| MoCo-v3Backbone=ViT-B, Evaluation Protocol=Linear Probing2024.04 | 76.7 | |
| DINOBackbone=ViT-B, Evaluation Protocol=K-Nearest Neighbor2024.04 | 76.1 | |
| I-JEPABackbone=ViT-B, Evaluation Protocol=Linear Probing2024.04 | 72.9 | |
| MAE (αmin = 0.6)Backbone=ViT-B, Evaluation Protocol=Linear Probing2024.04 | 72.7 | |
| Latent MIMBackbone=ViT-B, Evaluation Protocol=Linear Probing2024.04 | 72 | |
| CAEBackbone=ViT-B, Evaluation Protocol=Linear Probing2024.04 | 70.4 | |
| Data2VecBackbone=ViT-B, Evaluation Protocol=Linear Probing2024.04 | 68 | |
| MAEBackbone=ViT-B, Evaluation Protocol=Linear Probing2024.04 | 67.8 | |
| MoCo-v3Backbone=ViT-B, Evaluation Protocol=K-Nearest Neighbor2024.04 | 66.6 | |
| MAE (αmin = 0.6)Backbone=ViT-B, Evaluation Protocol=K-Nearest Neighbor2024.04 | 63.9 | |
| CAEBackbone=ViT-B, Evaluation Protocol=K-Nearest Neighbor2024.04 | 51.4 | |
| Latent MIMBackbone=ViT-B, Evaluation Protocol=K-Nearest Neighbor2024.04 | 50.1 | |
| Con MIMBackbone=ViT-B, Evaluation Protocol=Linear Probing2024.04 | 39.3 | |
| Data2VecBackbone=ViT-B, Evaluation Protocol=K-Nearest Neighbor2024.04 | 33.2 | |
| MAEBackbone=ViT-B, Evaluation Protocol=K-Nearest Neighbor2024.04 | 27.4 | |
| ADDPBackbone=VIT-L, Evaluation Protocol=Linear Probing2024.04 | 23.8 |