Auxiliary Image Regularization for Deep CNNs with Noisy Labels
This addresses the issue of noisy labels in image classification for real-world datasets with user-supplied tags, but it is incremental as it builds on existing regularization and optimization methods.
The paper tackles the problem of training deep CNNs with mislabeled training samples, common in real-world image datasets, by proposing an auxiliary image regularization technique optimized via stochastic ADMM, which exploits mutual context to select reliable images and robustify learning. Experiments on benchmark datasets show the model is resistant to label noise.
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.