Image Classification on Places205 (val)
64.35Top-1 AccuracySAMixP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SAMixPBackbone=ResNet-502021.11 | 64.35 | — | |
| SAMixBackbone=ResNet-502021.11 | 64.27 | — | |
| AutoMixBackbone=ResNet-502021.11 | 64.06 | — | |
| PuzzleMixBackbone=ResNet-502021.11 | 63.91 | — | |
| ResizeMixBackbone=ResNet-502021.11 | 63.88 | — | |
| CutMixBackbone=ResNet-502021.11 | 63.75 | — | |
| FMixBackbone=ResNet-502021.11 | 63.63 | — | |
| SaliencyMixBackbone=ResNet-502021.11 | 63.33 | — | |
| ManifoldMixBackbone=ResNet-502021.11 | 63.23 | — | |
| VanillaBackbone=ResNet-502021.11 | 63.1 | — | |
| MixupBackbone=ResNet-502021.11 | 63.01 | — | |
| SEER (RG-10B)Arch.=RG-10B, Pretrain=IG-1B, Param.=10B, Evaluation protocol=linear probe2022.02 | 62.9 | — | |
| SEER (RG-128Gf)Arch.=RG-128Gf, Pretrain=IG-1B, Param.=693M, Evaluation protocol=linear probe2022.02 | 61.9 | — | |
| Supervised ResNet50 v2Family=Residual, Protocol=Linear logistic regression2019.01 | 61.6 | 89 | |
| SwAV (RN50w5)Arch.=RN50w5, Pretrain=INet-1K, Param.=585M, Evaluation protocol=linear probe2022.02 | 60.3 | — | |
| SwAV (RG-128Gf)Arch.=RG-128Gf, Pretrain=INet-1K, Param.=693M, Evaluation protocol=linear probe2022.02 | 60.1 | — | |
| SAMixBackbone=ResNet-182021.11 | 59.86 | — | |
| SAMixPBackbone=ResNet-182021.11 | 59.82 | — | |
| AutoMixBackbone=ResNet-182021.11 | 59.74 | — | |
| ResizeMixBackbone=ResNet-182021.11 | 59.66 | — | |
| VanillaBackbone=ResNet-182021.11 | 59.63 | — | |
| PuzzleMixBackbone=ResNet-182021.11 | 59.62 | — | |
| FMixBackbone=ResNet-182021.11 | 59.51 | — | |
| SaliencyMixBackbone=ResNet-182021.11 | 59.5 | — | |
| ManifoldMixBackbone=ResNet-182021.11 | 59.46 | — | |
| MixupBackbone=ResNet-182021.11 | 59.33 | — | |
| CutMixBackbone=ResNet-182021.11 | 59.21 | — | |
| Supervised VGG19Family=VGG, Protocol=Linear logistic regression2019.01 | 58.9 | 89.3 | |
| Supervised RevNet50Family=Residual, Protocol=Linear logistic regression2019.01 | 58.9 | 87.5 | |
| DINO (ViT-B/8)Arch.=ViT-B/8, Pretrain=INet-1K, Param.=85M, Evaluation protocol=linear probe2022.02 | 57.7 | — | |
| OBoWEpochs=200, Batch=256, Evaluation Protocol=Linear probing, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 56.8 | — | |
| BYOLArch.=RN200w2, Pretrain=INet-1K, Param.=250M, Evaluation protocol=linear probe2022.02 | 56.8 | — | |
| SwAVEpochs=800, Batch=4096, Evaluation Protocol=Linear probing, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 56.5 | — | |
| SwAV (RN50)Arch.=RN50, Pretrain=INet-1K, Param.=25M, Evaluation protocol=linear probe2022.02 | 56.3 | — | |
| SwAVEpochs=200, Batch=256, Evaluation Protocol=Linear probing, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 56.2 | — | |
| Supervised (RG-128Gf)Arch.=RG-128Gf, Pretrain=INet-1K, Param.=693M, Evaluation protocol=linear probe2022.02 | 56 | — | |
| SCLRv2Arch.=RN152w3+SK, Pretrain=INet-1K, Param.=794M, Evaluation protocol=linear probe2022.02 | 56 | — | |
| DINO (RN50)Arch.=RN50, Pretrain=INet-1K, Param.=25M, Evaluation protocol=linear probe2022.02 | 55.9 | — | |
| DINO (ViT-B/16)Arch.=ViT-B/16, Pretrain=INet-1K, Param.=85M, Evaluation protocol=linear probe2022.02 | 55.2 | — | |
| AMDIMParameters=670M, Setup=Different architecture/evaluation setup2019.12 | 55.1 | — | |
| MoCov3Arch.=ViT-B/16, Pretrain=INet-1K, Param.=85M, Evaluation protocol=linear probe2022.02 | 53.9 | — | |
| Supervised (ViT-B/16)Arch.=ViT-B/16, Pretrain=INet-1K, Param.=85M, Evaluation protocol=linear probe2022.02 | 53.6 | — | |
| SimCLREpochs=1000, Batch=4096, Evaluation Protocol=Linear probing, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 53.3 | — | |
| SupervisedEpochs=100, Batch=256, Evaluation Protocol=Linear probing, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 53.2 | — | |
| MoCo v2Epochs=800, Batch=256, Evaluation Protocol=Linear probing, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 52.9 | — | |
| SupervisedBackbone=ResNet-50, Parameters=25.6M, Evaluation Setup=[19]2019.12 | 51.5 | — | |
| BoWNetEpochs=325, Batch=256, Evaluation Protocol=Linear probing, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 51.1 | — | |
| PIRLPre-training Dataset=YFCC1M, Backbone=ResNet-50, Evaluation Protocol=Linear classifier on fixed representations2019.12 | 51 | — | |
| PCLEpochs=200, Batch=256, Evaluation Protocol=Linear probing, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 50.3 | — | |
| LAParameters=25.6M, 10-crop evaluation=true, Setup=Different architecture/evaluation setup2019.12 | 50.2 | — | |
| PIRLBackbone=ResNet-50, Parameters=25.6M, Evaluation Setup=[19]2019.12 | 49.8 | — | |
| PIRLEpochs=800, Batch=1024, Evaluation Protocol=Linear probing, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 49.8 | — | |
| NPID++Backbone=ResNet-50, Parameters=25.6M, Evaluation Setup=[19]2019.12 | 46.4 | — | |
| NPIDParameters=25.6M, Setup=Different architecture/evaluation setup2019.12 | 45.5 | — | |
| Rot.Parameters=61M, Setup=Different architecture/evaluation setup2019.12 | 45.5 | — | |
| Rel. Patch Loc.Family=Residual, Protocol=Linear logistic regression2019.01 | 45.3 | 75.6 | |
| JigsawPre-training Dataset=YFCC100M, Backbone=ResNet-50, Evaluation Protocol=Linear classifier on fixed representations2019.12 | 44.8 | — | |
| ExemplarFamily=Residual, Protocol=Linear logistic regression2019.01 | 42.7 | 72.5 | |
| JigsawPre-training Dataset=YFCC1M, Backbone=ResNet-50, Evaluation Protocol=Linear classifier on fixed representations2019.12 | 42.1 | — | |
| DeeperClusterPre-training Dataset=YFCC100M, Backbone=VGG-16, Evaluation Protocol=Linear classifier on fixed representations2019.12 | 42.1 | — | |
| RotationBackbone=ResNet-50, Parameters=25.6M, Evaluation Setup=[19]2019.12 | 41.4 | — | |
| JigsawBackbone=ResNet-50, Parameters=25.6M, Evaluation Setup=[19]2019.12 | 41.2 | — | |
| ColorizationBackbone=ResNet-50, Parameters=25.6M, Evaluation Setup=[19]2019.12 | 37.5 | — | |
| DeepClusterParameters=61M, Setup=Different architecture/evaluation setup2019.12 | 37.5 | — | |
| AETParameters=61M, Setup=Different architecture/evaluation setup2019.12 | 37.1 | — | |
| JigsawFamily=AlexNet, Protocol=Linear logistic regression2019.01 | 35.5 | 71.6 | |
| DeepClusterPre-training Dataset=YFCC1M, Backbone=VGG-16, Evaluation Protocol=Linear classifier on fixed representations2019.12 | 35.4 | — | |
| RotationFamily=AlexNet, Protocol=Linear logistic regression2019.01 | 35.1 | 77.9 |