Improving the trustworthiness of image classification models by utilizing bounding-box annotations

TL;DR

This paper introduces a method that uses bounding-box annotations during training to enhance the accuracy, robustness, and interpretability of image classification models.

Contribution

It proposes a novel training objective that leverages bounding-box annotations to improve model trustworthiness in image classification tasks.

Findings

Improved accuracy over baseline models

Enhanced robustness to adversarial examples

Greater interpretability of model decisions

Abstract

We study utilizing auxiliary information in training data to improve the trustworthiness of machine learning models. Specifically, in the context of image classification, we propose to optimize a training objective that incorporates bounding box information, which is available in many image classification datasets. Preliminary experimental results show that the proposed algorithm achieves better performance in accuracy, robustness, and interpretability compared with baselines.