List your company
ResearchFormerly WizWand

Image Classification on CIFAR-100-N-LT synthetic (test)

48.98Average Accuracy (%)Dynamic Loss
24.467230.831137.19543.5589Nov 22, 2022
Updated 8mo ago

Evaluation results

MethodLinks
Dynamic LossSelection=Best, Backbone=PreAct ResNet-182022.1148.98—
Dynamic LossSelection=Last, Backbone=PreAct ResNet-182022.1148.56—
DivideMix + Balanced-SoftmaxSelection=Best, Backbone=PreAct ResNet-182022.1146.92—
DivideMix + Balanced-SoftmaxSelection=Last, Backbone=PreAct ResNet-182022.1146.11—
HARSelection=Best, Backbone=PreAct ResNet-182022.1143.98—
HARSelection=Last, Backbone=PreAct ResNet-182022.1142.77—
Balanced-SoftmaxSelection=Best, Backbone=PreAct ResNet-182022.1141.82—
DivideMixSelection=Best, Backbone=PreAct ResNet-182022.1141.75—
Balanced-SoftmaxSelection=Last, Backbone=PreAct ResNet-182022.1141.3—
DivideMixSelection=Last, Backbone=PreAct ResNet-182022.1141.15—
ELR+Selection=Best, Backbone=PreAct ResNet-182022.1136.2—
MOIT+Selection=Best, Backbone=PreAct ResNet-182022.1135.02—
ELR+Selection=Last, Backbone=PreAct ResNet-182022.1134.96—
CurveNetSelection=Best, Backbone=PreAct ResNet-182022.1132.72—
FaMUSSelection=Best, Backbone=PreAct ResNet-182022.1130.81—
FaMUSSelection=Last, Backbone=PreAct ResNet-182022.1130.72—
CurveNetSelection=Last, Backbone=PreAct ResNet-182022.1129.83—
Cross EntropySelection=Best, Backbone=PreAct ResNet-182022.1126.67—
Cross EntropySelection=Last, Backbone=PreAct ResNet-182022.1125.41—
Dynamic LossImbalance Ratio=10, Noise Rate=0.1, Selection=Best, Backbone=PreAct ResNet-182022.11—59.52
Dynamic LossImbalance Ratio=10, Noise Rate=0.5, Selection=Best, Backbone=PreAct ResNet-182022.11—51.26
Dynamic LossImbalance Ratio=100, Noise Rate=0.1, Selection=Best, Backbone=PreAct ResNet-182022.11—47.55
Dynamic LossImbalance Ratio=100, Noise Rate=0.5, Selection=Best, Backbone=PreAct ResNet-182022.11—34.3