Image Classification on Places205
71.7Top-1 AccuracyInternImage-H
Evaluation results
| Method | Links | ||
|---|---|---|---|
| InternImage-HBackbone=InternImage-H, Pre-training=427M joint dataset2022.11 | 71.7 | — | |
| OpenCLIPArch=ViT-G/14, Protocol=Frozen linear probe2023.04 | 69.8 | — | |
| MixMAEBackbone=Swin-L, FLOPs (G)=35.8, Params. (M)=2352022.05 | 69.3 | — | |
| MixMIM-LBackbone=MixMIM-L2022.11 | 69.3 | — | |
| MixMAEBackbone=Swin-B/W14, FLOPs (G)=16.3, Params. (M)=882022.05 | 68.6 | — | |
| DINOv2Arch=ViT-g/14, Protocol=Frozen linear probe2023.04 | 67.5 | — | |
| DINOv2Arch=ViT-L/14, Protocol=Frozen linear probe2023.04 | 67.3 | — | |
| MAEBackbone=ViT-H, Input Resolution=448, Pre-train Data=IN1K, Evaluation Protocol=Fine-tuned2021.11 | 66.8 | — | |
| DINOv2Arch=ViT-B/14, Protocol=Frozen linear probe2023.04 | 66.2 | — | |
| Previous bestPre-train Data=1 billion images2021.11 | 66 | — | |
| MAEBackbone=ViT-H, Pre-train Data=IN1K, Evaluation Protocol=Fine-tuned2021.11 | 65.9 | — | |
| MAEBackbone=ViT-L, Pre-train Data=IN1K, Evaluation Protocol=Fine-tuned2021.11 | 65.8 | — | |
| MAEBackbone=ViT-L, FLOPs (G)=61.3, Params. (M)=3042022.05 | 65.8 | — | |
| Ours (DINOv2)Data=INet-22k, Arch.=L/142024.06 | 65.1 | — | |
| iBOTArch=ViT-L/16, Protocol=Frozen linear probe2023.04 | 64.4 | — | |
| AutoMixBackbone=ResNet-502021.03 | 64.06 | — | |
| PuzzleMixBackbone=ResNet-502021.03 | 63.91 | — | |
| MAEBackbone=ViT-B, Pre-train Data=IN1K, Evaluation Protocol=Fine-tuned2021.11 | 63.9 | — | |
| MAEBackbone=ViT-B, FLOPs (G)=17.5, Params. (M)=862022.05 | 63.9 | — | |
| ResizeMix*Backbone=ResNet-502021.03 | 63.88 | — | |
| CutMixBackbone=ResNet-502021.03 | 63.75 | — | |
| FMix*Backbone=ResNet-502021.03 | 63.63 | — | |
| SaliencyMixBackbone=ResNet-502021.03 | 63.33 | — | |
| ManifoldMixBackbone=ResNet-502021.03 | 63.23 | — | |
| VanillaBackbone=ResNet-502021.03 | 63.1 | — | |
| MixUpBackbone=ResNet-502021.03 | 63.01 | — | |
| DINOv2Arch=ViT-S/14, Protocol=Frozen linear probe2023.04 | 62.9 | — | |
| DFNResolution=224, Pre-training Dataset=DFN-2B, Evaluation Protocol=Zero-shot2024.12 | 62 | — | |
| MetaCLIPResolution=224, Pre-training Dataset=MetaCLIP-2.5B, Evaluation Protocol=Zero-shot2024.12 | 61.8 | — | |
| EVA-02-CLIPResolution=336, Pre-training Dataset=Merged-2B, Evaluation Protocol=Zero-shot2024.12 | 61.7 | — | |
| dino.txtResolution=224, Pre-training Dataset=LVTD-2.3B, Evaluation Protocol=Zero-shot2024.12 | 61.2 | — | |
| dino.txtResolution=336, Pre-training Dataset=LVTD-2.3B, Evaluation Protocol=Zero-shot2024.12 | 61.2 | — | |
| DINOArch=ViT-B/8, Protocol=Frozen linear probe2023.04 | 60.4 | — | |
| AutoMixBackbone=ResNet-182021.03 | 59.74 | — | |
| ResizeMix*Backbone=ResNet-182021.03 | 59.66 | — | |
| VanillaBackbone=ResNet-182021.03 | 59.63 | — | |
| PuzzleMixBackbone=ResNet-182021.03 | 59.62 | — | |
| SigLIPResolution=384, Pre-training Dataset=WebLi, Evaluation Protocol=Zero-shot2024.12 | 59.6 | — | |
| FMix*Backbone=ResNet-182021.03 | 59.51 | — | |
| SaliencyMixBackbone=ResNet-182021.03 | 59.5 | — | |
| ManifoldMixBackbone=ResNet-182021.03 | 59.46 | — | |
| MixUpBackbone=ResNet-182021.03 | 59.33 | — | |
| CutMixBackbone=ResNet-182021.03 | 59.21 | — | |
| CLIPResolution=336, Pre-training Dataset=WIT-400M, Evaluation Protocol=Zero-shot2024.12 | 59.2 | — | |
| I-JEPAData=INet-22k, Arch.=g/162024.06 | 59.1 | — | |
| SigLIPResolution=256, Pre-training Dataset=WebLi, Evaluation Protocol=Zero-shot2024.12 | 58.6 | — | |
| I-JEPAData=INet-1k, Arch.=H/142024.06 | 58.4 | — | |
| BEITData=INet-1k, Arch.=L/162024.06 | 57.9 | — | |
| MAEData=INet-1k, Arch.=H/142024.06 | 56 | — | |
| SupervisedEvaluation Protocol=Linear evaluation2021.06 | 53.2 | — | |
| MAEArch=ViT-H/14, Protocol=Frozen linear probe2023.04 | 52.4 | — | |
| AIMData=DFN-2B+, Arch.=H/142024.06 | 51.7 | — | |
| Communication Game +augmentations +sharedEvaluation Protocol=Linear evaluation, Data Augmentations=True, Weight Sharing=True2021.06 | 49.1 | — | |
| SimCLREvaluation Protocol=Linear evaluation2021.06 | 49 | — | |
| Communication Game +augmentations -sharedEvaluation Protocol=Linear evaluation, Data Augmentations=True, Weight Sharing=False2021.06 | 47.9 | — |