DivideMix: Learning with Noisy Labels as Semi-supervised Learning

Benchmark Model Rank Results
learning-with-noisy-labels-on-cifar-100nDivide-Mix#4Accuracy (mean): 71.13
learning-with-noisy-labels-on-cifar-10n-worstDivide-Mix#6Accuracy (mean): 92.56
learning-with-noisy-labels-on-cifar-10n-aggregateDivide-Mix#7Accuracy (mean): 95.01
learning-with-noisy-labels-on-cifar-10n-random2Divide-Mix#8Accuracy (mean): 90.90
learning-with-noisy-labels-on-cifar-10n-random3Divide-Mix#12Accuracy (mean): 89.97
image-classification-on-clothing1mDivideMix#14Accuracy: 74.76%
learning-with-noisy-labels-on-cifar-10n-random1Divide-Mix#14Accuracy (mean): 90.18
image-classification-on-mini-webvision-1-0DivideMix (Inception-ResNet-v2)#28Top-1 Accuracy: 77.32Top-5 Accuracy: 91.64…
image-classification-on-mini-webvision-1-0DivideMix (ResNet-50)#30Top-1 Accuracy: 76.32 ±0.36Top-5 Accuracy: 90.65 ±0.16…
image-classification-on-mini-webvision-1-0DivideMix (ResNet-18)#31Top-1 Accuracy: 76.08