Confidence Calibration on iNaturalist 2021
0.65ECEPTSK + PROCAL
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| PTSK + PROCALBackbone=ResNet-50, Pre-training Dataset=ImageNet, Calibration Strategy=With PROCAL2023.06 | 0.65 | 0.82 | 11 | 1.41 | |
| ETS + PROCALBackbone=ResNet-50, Pre-training Dataset=ImageNet, Calibration Strategy=With PROCAL2023.06 | 0.66 | 0.77 | 13 | 1.38 | |
| TS + PROCALBackbone=ResNet-50, Pre-training Dataset=ImageNet, Calibration Strategy=With PROCAL2023.06 | 0.7 | 0.78 | 14 | 1.43 | |
| PTS + PROCALBackbone=ResNet-50, Pre-training Dataset=ImageNet, Calibration Strategy=With PROCAL2023.06 | 0.71 | 0.86 | 12 | 1.44 | |
| Conf + PROCALBackbone=ResNet-50, Pre-training Dataset=ImageNet, Calibration Strategy=With PROCAL2023.06 | 0.78 | 0.76 | 18 | 1.51 | |
| MIR + PROCALBackbone=ResNet-50, Pre-training Dataset=ImageNet, Calibration Strategy=With PROCAL2023.06 | 0.98 | 1.06 | 19 | 1.64 | |
| MIRBackbone=ResNet-50, Pre-training Dataset=ImageNet, Calibration Strategy=Base2023.06 | 1.05 | 1.09 | 18 | 1.63 | |
| ETSBackbone=ResNet-50, Pre-training Dataset=ImageNet, Calibration Strategy=Base2023.06 | 1.12 | 1.15 | 18 | 1.79 | |
| TSBackbone=ResNet-50, Pre-training Dataset=ImageNet, Calibration Strategy=Base2023.06 | 2.03 | 2.02 | 30 | 2.34 | |
| PTSKBackbone=ResNet-50, Pre-training Dataset=ImageNet, Calibration Strategy=Base2023.06 | 2.97 | 3.01 | 56 | 4.66 | |
| ConfBackbone=ResNet-50, Pre-training Dataset=ImageNet, Calibration Strategy=Base2023.06 | 4.85 | 4.86 | 55 | 4.91 | |
| PTSBackbone=ResNet-50, Pre-training Dataset=ImageNet, Calibration Strategy=Base2023.06 | 4.86 | 4.9 | 296 | 7.04 |