Image Classification on ImageNet 100% labels 1k
84.8Top-1 AccDeiT-III
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DeiT-IIIArchitecture=ViT-Huge, Param=632M2025.05 | 84.8 | — | |
| MoCov3Architecture=ViT-BN-L/7, Params.=304M, Epochs=300, Evaluation Protocol=Linear evaluation2022.04 | 81 | — | |
| iBOTBackbone=ViT-L/162026.04 | 81 | — | |
| MSNArchitecture=ViT-L/7, Params.=304M, Epochs=200, Evaluation Protocol=Linear evaluation2022.04 | 80.7 | — | |
| DINOArchitecture=ViT-B/8, Params.=86M, Epochs=300, Evaluation Protocol=Linear evaluation2022.04 | 80.1 | — | |
| DINOBackbone=ViT-B/82026.04 | 80.1 | — | |
| DeiTArchitecture=ViT-Small, Param=22M2025.05 | 79.8 | — | |
| SimCLRv2Architecture=RN151+SK (3×), Params.=795M, Epochs=800, Evaluation Protocol=Linear evaluation2022.04 | 79.8 | — | |
| BYOLArchitecture=RN200 (2×), Params.=250M, Epochs=800, Evaluation Protocol=Linear evaluation2022.04 | 79.6 | — | |
| iBOTArchitecture=ViT-B/16, Params.=86M, Epochs=400, Evaluation Protocol=Linear evaluation2022.04 | 79.4 | — | |
| SimCLRv2Backbone=RN152(2×)2026.04 | 79.1 | — | |
| ST-S+CBackbone=ResNet502026.04 | 78.4 | — | |
| iBOTArchitecture=ViT-S/16, Params.=22M, Epochs=800, Evaluation Protocol=Linear evaluation2022.04 | 77.9 | — | |
| DINOArchitecture=ViT-S/16, Params.=22M, Epochs=800, Evaluation Protocol=Linear evaluation2022.04 | 77 | — | |
| MSNArchitecture=ViT-S/16, Params.=22M, Epochs=600, Evaluation Protocol=Linear evaluation2022.04 | 76.9 | — | |
| MAEArchitecture=ViT-H/14, Params.=632M, Epochs=1600, Evaluation Protocol=Linear evaluation2022.04 | 76.6 | — | |
| ST-CBackbone=ResNet502026.04 | 75.4 | — | |
| BYOLArchitecture=RN50, Params.=24M, Epochs=1000, Evaluation Protocol=Linear evaluation2022.04 | 74.4 | — | |
| MoCo-v3Backbone=ResNet502026.04 | 73.8 | — | |
| SimCLRv2Architecture=RN50, Params.=24M, Epochs=800, Evaluation Protocol=Linear evaluation2022.04 | 71.7 | — | |
| MAE (public)Backbone=ViT-B/16, Evaluation Protocol=Fine-tuning with 100% labels2023.04 | — | 83.57 | |
| MAE (public)Backbone=ViT-L/16, Evaluation Protocol=Fine-tuning with 100% labels2023.04 | — | 85.87 | |
| MAE (reimpl)Backbone=ViT-B/16, Evaluation Protocol=Fine-tuning with 100% labels2023.04 | — | 83.2 | |
| MAE (reimpl)Backbone=ViT-L/16, Evaluation Protocol=Fine-tuning with 100% labels2023.04 | — | 85.37 | |
| MAE (reimpl)Backbone=ViT-H/16, Evaluation Protocol=Fine-tuning with 100% labels2023.04 | — | 86.5 | |
| MAE-CTaugBackbone=ViT-B/16, Evaluation Protocol=Fine-tuning with 100% labels2023.04 | — | 83.22 | |
| MAE-CTaugBackbone=ViT-L/16, Evaluation Protocol=Fine-tuning with 100% labels2023.04 | — | 85.56 | |
| MAE-CTaugBackbone=ViT-H/16, Evaluation Protocol=Fine-tuning with 100% labels2023.04 | — | 86.57 | |
| MAE-CTminBackbone=ViT-B/16, Evaluation Protocol=Fine-tuning with 100% labels2023.04 | — | 83.32 | |
| MAE-CTminBackbone=ViT-L/16, Evaluation Protocol=Fine-tuning with 100% labels2023.04 | — | 85.47 | |
| MAE-CTminBackbone=ViT-H/16, Evaluation Protocol=Fine-tuning with 100% labels2023.04 | — | 86.59 |