Auxiliary Image Regularization for Deep CNNs with Noisy Labels

TL;DR

This paper introduces an auxiliary image regularization method for deep CNNs that effectively handles noisy labels by leveraging mutual context among images, improving robustness in image classification tasks.

Contribution

The paper proposes a novel auxiliary regularization technique optimized with ADMM to automatically identify reliable images and mitigate label noise in CNN training.

Findings

The regularized CNN model shows increased resistance to label noise.

Experiments demonstrate improved accuracy on benchmark datasets with noisy labels.

The method effectively exploits mutual context among images to enhance learning robustness.

Abstract

Precisely-labeled data sets with sufficient amount of samples are very important for training deep convolutional neural networks (CNNs). However, many of the available real-world data sets contain erroneously labeled samples and those errors substantially hinder the learning of very accurate CNN models. In this work, we consider the problem of training a deep CNN model for image classification with mislabeled training samples - an issue that is common in real image data sets with tags supplied by amateur users. To solve this problem, we propose an auxiliary image regularization technique, optimized by the stochastic Alternating Direction Method of Multipliers (ADMM) algorithm, that automatically exploits the mutual context information among training images and encourages the model to select reliable images to robustify the learning process. Comprehensive experiments on benchmark data sets clearly demonstrate our proposed regularized CNN model is resistant to label noise in training data.