Learning with Instance-Dependent Noisy Labels by Anchor Hallucination and Hard Sample Label Correction

TL;DR

This paper introduces a novel approach for learning with instance-dependent noisy labels by hallucinating anchors and correcting hard sample labels, significantly improving accuracy on noisy datasets.

Contribution

It proposes a new method that explicitly distinguishes easy and hard samples, hallucinating anchors for label correction, and enhances learning from noisy, instance-dependent datasets.

Findings

Outperforms state-of-the-art NLL methods on synthetic datasets.

Effective in correcting labels of hard samples with complex visual patterns.

Improves semi-supervised training with corrected labels.

Abstract

Learning from noisy-labeled data is crucial for real-world applications. Traditional Noisy-Label Learning (NLL) methods categorize training data into clean and noisy sets based on the loss distribution of training samples. However, they often neglect that clean samples, especially those with intricate visual patterns, may also yield substantial losses. This oversight is particularly significant in datasets with Instance-Dependent Noise (IDN), where mislabeling probabilities correlate with visual appearance. Our approach explicitly distinguishes between clean vs.noisy and easy vs. hard samples. We identify training samples with small losses, assuming they have simple patterns and correct labels. Utilizing these easy samples, we hallucinate multiple anchors to select hard samples for label correction. Corrected hard samples, along with the easy samples, are used as labeled data in subsequent semi-supervised training. Experiments on synthetic and real-world IDN datasets demonstrate the superior performance of our method over other state-of-the-art NLL methods.