Learning with Instance-Dependent Label Noise: A Sample Sieve Approach

Benchmark Model Rank Results
image-classification-on-clothing1mCORES2#30Accuracy: 73.24%
learning-with-noisy-labels-on-cifar-100nCORES#9Accuracy (mean): 61.15
learning-with-noisy-labels-on-cifar-100nCORES*#21Accuracy (mean): 55.72
learning-with-noisy-labels-on-cifar-10nCORES*#6Accuracy (mean): 95.25
learning-with-noisy-labels-on-cifar-10nCORES#16Accuracy (mean): 91.23
learning-with-noisy-labels-on-cifar-10n-1CORES*#6Accuracy (mean): 94.45
learning-with-noisy-labels-on-cifar-10n-1CORES#17Accuracy (mean): 89.66
learning-with-noisy-labels-on-cifar-10n-2CORES*#4Accuracy (mean): 94.88
learning-with-noisy-labels-on-cifar-10n-2CORES#13Accuracy (mean): 89.91
learning-with-noisy-labels-on-cifar-10n-3CORES*#4Accuracy (mean): 94.74
learning-with-noisy-labels-on-cifar-10n-3CORES#13Accuracy (mean): 89.79
learning-with-noisy-labels-on-cifar-10n-worstCORES*#7Accuracy (mean): 91.66
learning-with-noisy-labels-on-cifar-10n-worstCORES#12Accuracy (mean): 83.60