Synergistic Network Learning and Label Correction for Noise-robust Image Classification

TL;DR

This paper introduces a robust label correction framework for noise-robust image classification that iteratively learns network parameters and reassigns labels, improving performance on noisy datasets.

Contribution

The proposed method combines small loss selection and noise correction, leveraging dual networks to iteratively identify and correct noisy labels in training data.

Findings

Outperforms baseline methods on CIFAR-10, CIFAR-100, Clothing1M.

Effectively identifies and corrects noisy labels across different noise types and rates.

Demonstrates robustness in real-world noisy datasets.

Abstract

Large training datasets almost always contain examples with inaccurate or incorrect labels. Deep Neural Networks (DNNs) tend to overfit training label noise, resulting in poorer model performance in practice. To address this problem, we propose a robust label correction framework combining the ideas of small loss selection and noise correction, which learns network parameters and reassigns ground truth labels iteratively. Taking the expertise of DNNs to learn meaningful patterns before fitting noise, our framework first trains two networks over the current dataset with small loss selection. Based on the classification loss and agreement loss of two networks, we can measure the confidence of training data. More and more confident samples are selected for label correction during the learning process. We demonstrate our method on both synthetic and real-world datasets with different noise types and rates, including CIFAR-10, CIFAR-100 and Clothing1M, where our method outperforms the baseline approaches.