Image Classification on ImageNet-1K 1.0 (1% labels)
80.2Top-1 AccDPT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DPTmode=ours2023.02 | 80.2 | 94 | — | |
| Semi-ViTSource=Reference [16]2023.02 | 80 | 93.1 | — | |
| Semi-ViTSource=Reproduced2023.02 | 79.4 | 93.4 | — | |
| DebiasPL w/ CLIP (multi-views)B.S.=768, epochs=50, Pre-train=true2022.01 | 70.9 | 89.3 | — | |
| DebiasPL w/ CLIPB.S.=384, epochs=50, Pre-train=true2022.01 | 69.1 | 89.1 | — | |
| DebiasPL (multi-views)B.S.=1536, epochs=300, Pre-train=true2022.01 | 67.1 | 85.8 | — | |
| CoMatch (multi-views)B.S.=1440, epochs=400, Pre-train=true2022.01 | 67.1 | 87.1 | — | |
| DebiasPL (multi-views)B.S.=768, epochs=200, Pre-train=true2022.01 | 66.5 | 85.6 | — | |
| PAWS (multi-crops)B.S.=4096, epochs=50, Pre-train=false2022.01 | 66.5 | — | — | |
| DebiasPL (multi-views)B.S.=768, epochs=50, Pre-train=true2022.01 | 65.3 | 85.2 | — | |
| SMoGBackbone=resnet-50 x22022.07 | 63.6 | 85.6 | — | |
| DebiasPL w/ FixMatchB.S.=384, epochs=50, Pre-train=true2022.01 | 63.1 | 83.6 | — | |
| BYOLBackbone=resnet-50 x22022.07 | 62.2 | 84.1 | — | |
| FixMatch w/ EMANB.S.=384, epochs=50, Pre-train=true2022.01 | 60.9 | 82.5 | — | |
| SimCLRv2 (+ Self-distillation)B.S.=4096, epochs=400, Pre-train=true2022.01 | 60 | 79.8 | — | |
| FixMatch w/ DAB.S.=4096, epochs=400, Pre-train=true2022.01 | 59.9 | 79.8 | — | |
| SimCLRBackbone=resnet-50 x22022.07 | 58.5 | 83 | — | |
| SMoGBackbone=resnet-50 x12022.07 | 58 | 81.6 | — | |
| NNCLRBackbone=resnet-50 x12022.07 | 56.4 | 80.7 | — | |
| Barlow TwinsBackbone=resnet-50 x12022.07 | 55 | 79.2 | — | |
| SWAVB.S.=4096, epochs=50, Pre-train=true2022.01 | 53.9 | 78.5 | — | |
| SWAVBackbone=resnet-50 x12022.07 | 53.9 | 78.5 | — | |
| FixMatch w/ DAB.S.=4096, epochs=400, Pre-train=false2022.01 | 53.4 | 74.4 | — | |
| CLIP (few-shot)B.S.=256, epochs=50, Pre-train=true2022.01 | 53.4 | — | — | |
| BYOLBackbone=resnet-50 x12022.07 | 53.2 | 78.4 | — | |
| SSL-HSICBackbone=resnet-50 x12022.07 | 52.1 | 77.7 | — | |
| SimCLRBackbone=resnet-50 x12022.07 | 48.3 | 75.5 | — | |
| SupervisedBackbone=resnet-50 x12022.07 | 25.4 | 48.4 | — | |
| CLIP (Language Init.)Backbone=ViT-B/16, # Params=86M, Initialization=Language Init.2023.01 | — | — | 74.3 | |
| CLIP (Language Init.)Backbone=ViT-L/14, # Params=304M, Initialization=Language Init.2023.01 | — | — | 80.5 | |
| CLIP (Random Init.)Backbone=ViT-B/16, # Params=86M, Initialization=Random Init.2023.01 | — | — | 70.9 | |
| DINOBackbone=ViT-B/8, # Params=86M2023.01 | — | — | 70 | |
| iBOTBackbone=ViT-B/16, # Params=86M2023.01 | — | — | 69.7 | |
| MSNBackbone=ViT-B/4, # Params=86M2023.01 | — | — | 75.7 | |
| MSNBackbone=ViT-L/7, # Params=304M2023.01 | — | — | 75.1 | |
| PAWSBackbone=RN50x4, # Params=375M2023.01 | — | — | 69.9 | |
| REACT (Locked-Text Gated-Image)Backbone=ViT-B/16, # Params=129M, Initialization=Locked-Text Gated-Image2023.01 | — | — | 77.4 | |
| REACT (Locked-Text Gated-Image)Backbone=ViT-L/14, # Params=380M, Initialization=Locked-Text Gated-Image2023.01 | — | — | 81.6 | |
| REACT (Locked-Text)Backbone=ViT-B/16, # Params=86M, Initialization=Locked-Text2023.01 | — | — | 76.1 | |
| REACT (Locked-Text)Backbone=ViT-L/14, # Params=304M, Initialization=Locked-Text2023.01 | — | — | 81.6 | |
| SEERBackbone=RegNetY, # Params=1.3B2023.01 | — | — | 60.5 | |
| Semi-ViTBackbone=ViT-H/14, # Params=632M2023.01 | — | — | 80 | |
| SimCLRv2Backbone=RN152x3, # Params=795M2023.01 | — | — | 74.9 | |
| SimCLRv2 (self-distilled)Backbone=RN152x3, # Params=795M2023.01 | — | — | 76.6 |