Difficulty-Net: Learning to Predict Difficulty for Long-Tailed Recognition

Open paper
Benchmark Model Rank Results
long-tail-learning-on-places-ltDifficulty-Net (ResNet-152)#11Top-1 Accuracy: 41.7
long-tail-learning-on-cifar-100-lt-r-10Difficulty-Net#12Error Rate: 34.78
long-tail-learning-on-cifar-100-lt-r-50Difficulty-Net#15Error Rate: 43.1
long-tail-learning-on-cifar-100-lt-r-100Difficulty-Net#21Error Rate: 47.04
long-tail-learning-on-imagenet-ltDifficulty-Net (ResNet-50 using RandAugment, single model)#31Top-1 Accuracy: 57.4
long-tail-learning-on-imagenet-ltDifficulty-Net (ResNet-50 w/o using RandAugment, single model)#41Top-1 Accuracy: 54.0
long-tail-learning-on-imagenet-ltDifficulty-Net (ResNet-10 w/o using RandAugment, single model#57Top-1 Accuracy: 44.6

Rank counts only results with a code link.