Image Classification on ImageNet-C (AUROC and Indexed Accuracy/ECE)
94.67AUROCCSAM
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| CSAMBackbone=ResNet-50, Optimizer=CSAM2025.05 | 94.67 | 38.29 | 37.11 | 28.69 | 3.28 | 3.02 | 5.47 | |
| SAMBackbone=ResNet-50, Optimizer=SAM2025.05 | 94.35 | 37.45 | 36.35 | 27.85 | 4.91 | 3.74 | 6.92 | |
| SGDBackbone=ResNet-50, Optimizer=SGD2025.05 | 94.01 | 36.89 | 35.81 | 24.99 | 7.97 | 4.23 | 17.29 | |
| CSAMBackbone=ViT-S/16, Optimizer=CSAM2025.05 | 93.66 | 44.01 | 39.82 | 29.57 | 2.81 | 3.24 | 4.75 | |
| SAMBackbone=ViT-S/16, Optimizer=SAM2025.05 | 93.27 | 43.36 | 39.15 | 28.93 | 2.92 | 3.58 | 5.02 | |
| CSAMBackbone=ViT-S/32, Optimizer=CSAM2025.05 | 91.54 | 38.88 | 36.94 | 34.16 | 3.01 | 5.76 | 5.41 | |
| SAMBackbone=ViT-S/32, Optimizer=SAM2025.05 | 91.01 | 37.95 | 36.09 | 33.36 | 3.35 | 6.27 | 7.89 | |
| AdamWBackbone=ViT-S/16, Optimizer=AdamW2025.05 | 90.61 | 37.4 | 35.54 | 24.26 | 14.14 | 12.27 | 18.63 | |
| AdamWBackbone=ViT-S/32, Optimizer=AdamW2025.05 | 88.63 | 33.53 | 32.87 | 26.48 | 14.73 | 12.28 | 19.57 |