🤖 AI Summary
To address the challenge of identifying and cleansing “hard noise” in long-tailed distributions, this paper introduces the novel problem of Noisy Long-Tailed Classification (NLT). We propose an iterative Hard-to-Easy (H2E) self-bootstrapping denoising framework: a distribution-invariant noise detector dynamically assesses noise hardness, while context-agnostic feature disentanglement, multi-stage reweighting, and progressive cleaning transform hard noise into easy noise. Our method jointly enhances classifier robustness and denoising accuracy on both long-tailed and balanced data. Extensive experiments on three benchmarks—ImageNet-NLT, Animal10-NLT, and Food101-NLT—demonstrate substantial improvements over state-of-the-art methods: +5.2% average accuracy under long-tailed settings, with no degradation in balanced-scenario performance.
📝 Abstract
Conventional de-noising methods rely on the assumption that all samples are independent and identically distributed, so the resultant classifier, though disturbed by noise, can still easily identify the noises as the outliers of training distribution. However, the assumption is unrealistic in large-scale data that is inevitably long-tailed. Such imbalanced training data makes a classifier less discriminative for the tail classes, whose previously"easy"noises are now turned into"hard"ones -- they are almost as outliers as the clean tail samples. We introduce this new challenge as Noisy Long-Tailed Classification (NLT). Not surprisingly, we find that most de-noising methods fail to identify the hard noises, resulting in significant performance drop on the three proposed NLT benchmarks: ImageNet-NLT, Animal10-NLT, and Food101-NLT. To this end, we design an iterative noisy learning framework called Hard-to-Easy (H2E). Our bootstrapping philosophy is to first learn a classifier as noise identifier invariant to the class and context distributional changes, reducing"hard"noises to"easy"ones, whose removal further improves the invariance. Experimental results show that our H2E outperforms state-of-the-art de-noising methods and their ablations on long-tailed settings while maintaining a stable performance on the conventional balanced settings. Datasets and codes are available at https://github.com/yxymessi/H2E-Framework
