Fine-grained Recognition on iNaturalist 2018 (test)
84.7Top-1 AccSEER
Evaluation results
| Method | Links | |
|---|---|---|
| SEERArch.=RG-10B, Pretrain=IG-1B, Evaluation Protocol=finetuned2022.02 | 84.7 | |
| SEERArch.=RG-128Gf, Pretrain=IG-1B, Evaluation Protocol=finetuned2022.02 | 82.6 | |
| SupervisedArch.=ViT-L/16, Pretrain=INet-1K, Evaluation Protocol=finetuned2022.02 | 81.7 | |
| SupervisedArch.=ViT-B/16, Pretrain=INet-1K, Evaluation Protocol=finetuned2022.02 | 79.8 | |
| SwAVArch.=RG-128Gf, Pretrain=INet-1K, Evaluation Protocol=finetuned2022.02 | 79.7 | |
| DeiTArch.=ViT-B/16, Pretrain=INet-1K, Evaluation Protocol=finetuned2022.02 | 79.5 | |
| SupervisedArch.=RG-128Gf, Pretrain=INet-1K, Evaluation Protocol=finetuned2022.02 | 78.7 | |
| BYOLArch.=RN200w2, Pretrain=INet-1K, Evaluation Protocol=finetuned2022.02 | 76.1 | |
| SwAVArch.=RN50w5, Pretrain=INet-1K, Evaluation Protocol=finetuned2022.02 | 76 | |
| SCLRv2Arch.=RN152w3+SK, Pretrain=INet-1K, Evaluation Protocol=finetuned2022.02 | 74.1 | |
| DINOArch.=ViT-B/16, Pretrain=INet-1K, Evaluation Protocol=finetuned2022.02 | 72.6 | |
| SEERArch.=RG-10B, Pretrain=IG-1B, Evaluation Protocol=linear2022.02 | 53 | |
| DINOArch.=ViT-B/16, Pretrain=INet-1K, Evaluation Protocol=linear2022.02 | 50.1 | |
| SwAVArch.=RN50w5, Pretrain=INet-1K, Evaluation Protocol=linear2022.02 | 48.6 | |
| SwAVArch.=RG-128Gf, Pretrain=INet-1K, Evaluation Protocol=linear2022.02 | 47.5 | |
| SupervisedArch.=RG-128Gf, Pretrain=INet-1K, Evaluation Protocol=linear2022.02 | 47.2 | |
| SEERArch.=RG-128Gf, Pretrain=IG-1B, Evaluation Protocol=linear2022.02 | 47.2 | |
| BYOLArch.=RN200w2, Pretrain=INet-1K, Evaluation Protocol=linear2022.02 | 45.7 | |
| SCLRv2Arch.=RN152w3+SK, Pretrain=INet-1K, Evaluation Protocol=linear2022.02 | 43 | |
| SupervisedArch.=ViT-B/16, Pretrain=INet-1K, Evaluation Protocol=linear2022.02 | 40.7 |