CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features

Open paper
Benchmark Model Rank Results
domain-generalization-on-imagenet-aCutMix (ResNet-50)#33Top-1 accuracy %: 7.3
image-captioning-on-cocoNIC (ResNet-50, CutMix)#5CIDEr: 77.6BLEU-1: 64.2BLEU-2: 46.3BLEU-3: 33.6BLEU-4: 24.9…
image-classification-on-cifar-10PyramidNet-200 + CutMix#91Percentage correct: 97.12
image-classification-on-cifar-100PyramidNet-200 + Shakedrop + Cutmix#55Percentage correct: 86.19
image-classification-on-imagenetResNeXt-101 (CutMix)#663Top 1 Accuracy: 80.53%
image-classification-on-imagenetResNet-50 (CutMix)#787Top 1 Accuracy: 78.4%
image-classification-on-omnibenchmarkCutMix#19Average Top-1 Accuracy: 31.1
semantic-segmentation-on-acdc-scribblesCutMix#4Dice (Average): 70.5%

Rank counts only results with a code link.