List your company
ResearchFormerly WizWand

Image Classification on ObjectNet-9 distribution-shifted (test)

31.38Test AccuracyBYOL + BG-Mixup (CAM)
21.853624.326826.829.2732Jul 30, 2021
Updated 8mo ago

Evaluation results

MethodLinks
BYOL + BG-Mixup (CAM)Augmentation=+BG-Mixup (CAM), Evaluation protocol=Linear classifier, Training dataset=ImageNet-92021.0731.38
BYOL + BG-HardMix (GT)Augmentation=+BG-HardMix (GT), Evaluation protocol=Linear classifier, Training dataset=ImageNet-92021.0730.78
MoCov2 + BG-Mixup (CAM)Augmentation=+BG-Mixup (CAM), Evaluation protocol=Linear classifier, Training dataset=ImageNet-92021.0729.28
MoCov2 + MixupAugmentation=Mixup [41], Evaluation protocol=Linear classifier, Training dataset=ImageNet-92021.0728.53
MoCov2 + BG-HardMix (GT)Augmentation=+BG-HardMix (GT), Evaluation protocol=Linear classifier, Training dataset=ImageNet-92021.0727.48
MoCov2 + CutMixAugmentation=CutMix [42], Evaluation protocol=Linear classifier, Training dataset=ImageNet-92021.0726.87
BYOL + MixupAugmentation=Mixup [41], Evaluation protocol=Linear classifier, Training dataset=ImageNet-92021.0725.97
BYOLAugmentation=Baseline, Evaluation protocol=Linear classifier, Training dataset=ImageNet-92021.0725.38
MoCov2Augmentation=Baseline, Evaluation protocol=Linear classifier, Training dataset=ImageNet-92021.0724.96
BYOL + CutMixAugmentation=CutMix [42], Evaluation protocol=Linear classifier, Training dataset=ImageNet-92021.0722.22