Probability Calibration on ImageNet-16H Noise Level 110
3.92ECEP+L
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| P+LBackbone=VGG-19, Calibration fitting data size=5000, Evaluator type=Recalibrated Combination2021.09 | 3.92 | 0.96 | 0.41 | |
| P+LBackbone=VGG-19, Calibration fitting data size=5000, Evaluator type=Combination2021.09 | 6.07 | 1.03 | 0.49 | |
| P+LBackbone=VGG-19, Calibration fitting data size=10, Evaluator type=Combination2021.09 | 6.31 | 1.04 | 0.51 | |
| VGG-19Backbone=VGG-19, Calibration fitting data size=5000, Evaluator type=Model2021.09 | 7.24 | 1.36 | 0.57 | |
| P+LBackbone=VGG-19, Calibration fitting data size=0, Evaluator type=Combination2021.09 | 7.81 | 1.14 | 0.61 | |
| VGG-19Backbone=VGG-19, Calibration fitting data size=10, Evaluator type=Model2021.09 | 7.81 | 1.38 | 0.6 | |
| VGG-19Backbone=VGG-19, Calibration fitting data size=0, Evaluator type=Model2021.09 | 9.76 | 1.44 | 0.73 |