Image Classification on CIFAR-10 (Uncertainty & Calibration)
80.3AccuracyCVP
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| CVPRel. Time=2.2×, Backbone=sub-tiny ViT2026.06 | 80.3 | 0.0552 | 20.2 | 28.9 | 0.56 | 0.91 | 0.28 | |
| Mean NetworkRel. Time=0.9×, Backbone=sub-tiny ViT2026.06 | 79.9 | 0.0584 | 16.8 | 26.6 | 0.63 | 6.92 | 0.29 | |
| Mean Network + Temp. Sc.Rel. Time=0.9×, Backbone=sub-tiny ViT2026.06 | 79.9 | 0.0578 | 16.7 | 27.7 | 0.58 | 1.11 | 0.28 | |
| StreamliningRel. Time=2.0×, Backbone=sub-tiny ViT2026.06 | 79.8 | 0.0571 | 19.9 | 28.9 | 0.58 | 1.06 | 0.28 | |
| MC Sampling (4 samples)Rel. Time=4.0×, Backbone=sub-tiny ViT2026.06 | 79.3 | 0.0614 | 16.1 | 25.2 | 0.61 | 1.21 | 0.29 | |
| MC Sampling (2 samples)Rel. Time=2.0×, Backbone=sub-tiny ViT2026.06 | 78.3 | 0.067 | 15.3 | 22.6 | 0.64 | 2.79 | 0.31 |