Image Classification on ImageNet-1K (val) (Certified Robustness Evaluation)
83.8Empirical AccuracyDiffusion Calibration
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Diffusion Calibrationsigma=(0.25-0.50), Backbone=ViT-B/16@2242023.10 | 83.8 | 76.6 | 54.6 | 39.8 | 23 | — | 0.743 | |
| Cascaded Randomized Smoothingsigma=(0.25-0.50), Backbone=ViT-B/16@2242023.10 | 83.5 | 75 | 52.6 | 39 | 22.8 | — | 0.72 | |
| Diffusion Calibrationsigma=(0.25-1.00), Backbone=ViT-B/16@2242023.10 | 83.2 | 72.5 | 44 | 27.5 | 19.9 | 14.1 | 0.775 | |
| Cascaded Randomized Smoothingsigma=(0.25-1.00), Backbone=ViT-B/16@2242023.10 | 82.6 | 69 | 42.4 | 26.6 | 19 | 14.6 | 0.752 | |
| Carlini et al.sigma=0.25, Backbone=ViT-B/16@2242023.10 | 80.9 | 78.8 | 61.4 | — | — | — | 0.517 | |
| Carlini et al.sigma=0.50, Backbone=ViT-B/16@2242023.10 | 79.5 | 65.8 | 52.2 | 38.6 | 23.6 | — | 0.703 | |
| Carlini et al.sigma=1.00, Backbone=ViT-B/16@2242023.10 | 77.2 | 30.6 | 25.8 | 20.6 | 17 | 12.6 | 0.538 |