Uncertainty Based Detection and Relabeling of Noisy Image Labels

TL;DR

This paper introduces a method that uses predictive uncertainty to detect and relabel noisy labels in training data, improving the robustness of deep neural networks in noisy label scenarios.

Contribution

The paper presents a novel uncertainty-based approach for identifying and relabeling noisy data in training sets, enhancing DNN generalization.

Findings

Effective detection of noisy labels on CIFAR-10 and CIFAR-100.

Iterative relabeling improves data quality and model performance.

Method is simple to implement and adaptable.

Abstract

Deep neural networks (DNNs) are powerful tools in computer vision tasks. However, in many realistic scenarios label noise is prevalent in the training images, and overfitting to these noisy labels can significantly harm the generalization performance of DNNs. We propose a novel technique to identify data with noisy labels based on the different distributions of the predictive uncertainties from a DNN over the clean and noisy data. Additionally, the behavior of the uncertainty over the course of training helps to identify the network weights which best can be used to relabel the noisy labels. Data with noisy labels can therefore be cleaned in an iterative process. Our proposed method can be easily implemented, and shows promising performance on the task of noisy label detection on CIFAR-10 and CIFAR-100.