Image Classification on ImageNet-R (val)
96.5AccuracyCoCa
Evaluation Results
| Method | Links | |
|---|---|---|
| CoCaEvaluation Protocol=Zero-shot2023.02 | 96.5 | |
| LiT-e/14Evaluation Protocol=Zero-shot, Backbone=ViT-e/142023.02 | 96.1 | |
| LiT-22BEvaluation Protocol=Zero-shot, Backbone=ViT-22B/142023.02 | 96 | |
| BASICEvaluation Protocol=Zero-shot2023.02 | 95.7 | |
| LiT-g/14Evaluation Protocol=Zero-shot, Backbone=ViT-g/142023.02 | 94.9 | |
| ALIGNEvaluation Protocol=Zero-shot2023.02 | 92.2 | |
| CLIPEvaluation Protocol=Zero-shot2023.02 | 88.9 | |
| ZEROBackbone=CLIP-ViT-B-16 + CLIP-ViT-L-142024.05 | 87.21 | |
| RLCFBackbone=CLIP-ViT-B-16 + CLIP-ViT-L-14, Tuning=full image encoder tuning (Theta_i), Adaptation steps (t)=32024.05 | 86.19 | |
| Zero-ShotBackbone=CLIP-ViT-B-16 + CLIP-ViT-L-142024.05 | 85.4 | |
| RLCFBackbone=CLIP-ViT-B-16 + CLIP-ViT-L-14, Tuning=prompt tuning (Theta_etx), Adaptation steps (t)=12024.05 | 83.35 | |
| PaLI-17BShot Setting=0-shot2022.09 | 81.97 | |
| PaLI-15BShot Setting=0-shot2022.09 | 81.21 | |
| ZEROBackbone=CLIP-ViT-B-16, Ensemble variant=true2024.05 | 80.75 | |
| PaLI-3BShot Setting=0-shot2022.09 | 80.15 | |
| ZEROBackbone=MaPLe2024.05 | 79.74 | |
| PromptAlignBackbone=MaPLe2024.05 | 79.33 | |
| TPTBackbone=MaPLe2024.05 | 78.12 | |
| EnsembleBackbone=CLIP-ViT-B-162024.05 | 77.65 | |
| ZEROBackbone=CLIP-ViT-B-162024.05 | 77.22 | |
| TPTBackbone=CLIP-ViT-B-162024.05 | 77.06 | |
| Zero-ShotBackbone=MaPLe2024.05 | 76.98 | |
| Zero-ShotBackbone=CLIP-ViT-B-162024.05 | 73.98 | |
| Zero-Shot baselineEvaluation Protocol=Zero-Shot, Label Distribution Estimation=None2024.04 | 68.9 | |
| OTTEREvaluation Protocol=Zero-Shot, Label Distribution Estimation=BBSE+OT2024.04 | 63.3 | |
| OTTEREvaluation Protocol=Linear Probing, Label Distribution Estimation=BBSE+OT2024.04 | 57.1 | |
| STL (FAN-L-Hybrid)Params (M)=77.32024.01 | 56.6 | |
| STL (FAN-B-Hybrid)Params (M)=50.92024.01 | 55.3 | |
| Linear Probing baselineEvaluation Protocol=Linear Probing, Label Distribution Estimation=None2024.04 | 54.9 | |
| FAN-L-HybridParams (M)=76.82024.01 | 53.2 | |
| FAN-B-HybridParams (M)=50.42024.01 | 52.9 | |
| STL (FAN-S-Hybrid)Params (M)=26.52024.01 | 51.8 | |
| ConvNext-BParams (M)=88.62024.01 | 51.3 | |
| FAN-S-HybridParams (M)=26.32024.01 | 50.7 | |
| MSNArchitecture=ViT-B/16, Pre-training method=MSN2022.04 | 50 | |
| ConvNext-SParams (M)=50.22024.01 | 49.5 | |
| RVT-BParams (M)=91.82024.01 | 48.7 | |
| MAE-ViT-BParams (M)=862024.01 | 48.3 | |
| MAEArchitecture=ViT-B/16, Pre-training method=MAE2022.04 | 48.3 | |
| RVT-SParams (M)=23.32024.01 | 47.7 | |
| BBSE+PMEvaluation Protocol=Linear Probing, Label Distribution Estimation=BBSE+PM2024.04 | 47.6 | |
| DAT-AugReg-ViTParams (M)=862024.01 | 47.3 | |
| ConvNext-TParams (M)=28.62024.01 | 47.2 | |
| LV-ViT-MParams (M)=562024.01 | 47.2 | |
| Swin-SParams (M)=502024.01 | 46.6 | |
| Swin-BParams (M)=87.82024.01 | 46.6 | |
| LV-ViT-SParams (M)=262024.01 | 45.8 | |
| XCIT-S12Params (M)=26.32024.01 | 45.5 | |
| XCIT-S24Params (M)=47.72024.01 | 45.5 | |
| CvT*-LS-21Params (M)=32.12021.07 | 45 | |
| LSNet-BFLOPS=1.3G2025.03 | 43.1 | |
| CvT-13Params (M)=202021.07 | 42.9 | |
| CvT-21Params (M)=322021.07 | 42.6 | |
| CvT*-LS-13Params (M)=20.32021.07 | 42.6 | |
| DeiT-SParams (M)=22.12021.07 | 41.9 | |
| PVTv2-B1FLOPS=2.1G2025.03 | 41.8 | |
| Swin-TParams (M)=28.32024.01 | 41.3 | |
| FasterNet-T2FLOPS=1.9G2025.03 | 40.5 | |
| EdgeNeXt-SFLOPS=1.3G2025.03 | 40.1 | |
| FastViT-T12FLOPS=1.4G2025.03 | 39.9 | |
| LSNet-SFLOPS=0.5G2025.03 | 39.4 | |
| LSNet-TFLOPS=0.3G2025.03 | 38.5 | |
| UniRepLKNet-AFLOPS=0.6G2025.03 | 37.9 | |
| PoolFormer-S12FLOPS=1.8G2025.03 | 37.7 | |
| FastViT-T8FLOPS=0.7G2025.03 | 36.8 | |
| ResNet50Architecture=ResNet50, Pre-training method=Supervised2022.04 | 36.11 | |
| EfficientViT-M3FLOPS=0.3G2025.03 | 36.1 | |
| ResNet-50Params (M)=25.62021.07 | 35.3 | |
| PVTv2-B0FLOPS=0.6G2025.03 | 34.2 | |
| StarNet-S1FLOPS=0.4G2025.03 | 34.1 | |
| PVT-TinyFLOPS=1.9G2025.03 | 33.9 | |
| EdgeNeXt-XSFLOPS=0.5G2025.03 | 32.5 | |
| EdgeNeXt-XXSFLOPS=0.3G2025.03 | 29.5 | |
| MoCo-v2 + Patch-base NSalpha=2, k=163842021.10 | 29.45 | |
| MoCo-v2 + Patch-base NSalpha=2, k=655362021.10 | 28.95 | |
| FasterNet-T0FLOPS=0.3G2025.03 | 28.6 | |
| MoCo-v2 + Patch-base NSalpha=2, k=327682021.10 | 28.5 | |
| MoCo-v2 + Patch-base NSalpha=3, k=655362021.10 | 28.11 | |
| MoCo-v22021.10 | 27.81 | |
| MoCo-v2 + MoCHiN=512, K=1024, s=5122021.10 | 25.71 | |
| MoCo-v2 + MoCHiN=128, K=1024, s=5122021.10 | 25.29 | |
| BBSE+PMEvaluation Protocol=Zero-Shot, Label Distribution Estimation=BBSE+PM2024.04 | 1.7 |