List your company
ResearchFormerly WizWand

Selective Classification on Tiny-ImageNet (test)

76.64AUCOCCE+AUCOCL
73.925674.630375.33576.0397Aug 9, 2023
Updated 8mo ago

Evaluation results

MethodLinks
CE+AUCOCLLoss function=CE+AUCOCLoss, Backbone=WideResnet-28-102023.0876.6451.429.9747.65
FL+AUCOCLLoss function=FL+AUCOCLoss, Backbone=WideResnet-28-102023.0875.8650.843250.12
CE+S-ECELoss function=CE+S-ECE, Backbone=WideResnet-28-102023.0875.6249.9131.8849.55
AdaFL53Loss function=AdaFL53, Backbone=WideResnet-28-102023.0875.3649.4933.7750.76
FL (γ=3)Loss function=Focal Loss, Backbone=WideResnet-28-102023.0875.0549.5634.1751.72
FL+MMCELoss function=FL+MMCE, Backbone=WideResnet-28-102023.0874.7949.3234.5752.12
FL+S-ECELoss function=FL+S-ECE, Backbone=WideResnet-28-102023.0874.5749.1635.652.36
FL+ S-AvUCLoss function=FL+ S-AvUC, Backbone=WideResnet-28-102023.0874.2149.0235.2253.12
CE+S-AvUCLoss function=CE+S-AvUC, Backbone=WideResnet-28-102023.0874.1449.133553.17
CE+MMCELoss function=CE+MMCE, Backbone=WideResnet-28-102023.0874.1248.5635.8352.48
CELoss function=Cross-Entropy, Backbone=WideResnet-28-102023.0874.0349.2934.7452.28