Deep Learning Classifiers for Automated Driving: Quantifying the Trained DNN Model’s Vulnerability to Misclassification

Himanshu Agarwal, Himanshu Agarwal, Rafal Dorociak, Achim Rettberg, Achim Rettberg

2021

Abstract

The perception-based tasks in automated driving depend greatly on deep neural networks (DNNs). In context of image classification, the identification of the critical pairs of the target classes that make the DNN highly vulnerable to misclassification can serve as a preliminary step before implementing the appropriate measures for improving the robustness of the DNNs or the classification functionality. In this paper, we propose that the DNN’s vulnerability to misclassifying an input image into a particular incorrect class can be quantified by evaluating the similarity learnt by the trained model between the true class and the incorrect class. We also present the criteria to rank the DNN model’s vulnerability to a particular misclassification as either low, moderate or high. To argue for the validity of our proposal, we conduct an empirical assessment on DNN-based traffic sign classification. We show that upon evaluating the DNN model, most of the images for which it yields an erroneous prediction experience the misclassifications to which its vulnerability was ranked as high. Furthermore, we also validate empirically that all those possible misclassifications to which the DNN model’s vulnerability is ranked as high are difficult to deal with or control, as compared to the other possible misclassifications.

Paper Citation

in Harvard Style

Agarwal H., Dorociak R. and Rettberg A. (2021). Deep Learning Classifiers for Automated Driving: Quantifying the Trained DNN Model’s Vulnerability to Misclassification. In Proceedings of the 7th International Conference on Vehicle Technology and Intelligent Transport Systems - Volume 1: VEHITS, ISBN 978-989-758-513-5, pages 211-222. DOI: 10.5220/0010481302110222

in Bibtex Style

@conference{vehits21,
author={Himanshu Agarwal and Rafal Dorociak and Achim Rettberg},
title={Deep Learning Classifiers for Automated Driving: Quantifying the Trained DNN Model’s Vulnerability to Misclassification},
booktitle={Proceedings of the 7th International Conference on Vehicle Technology and Intelligent Transport Systems - Volume 1: VEHITS,},
year={2021},
pages={211-222},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0010481302110222},
isbn={978-989-758-513-5},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 7th International Conference on Vehicle Technology and Intelligent Transport Systems - Volume 1: VEHITS,
TI - Deep Learning Classifiers for Automated Driving: Quantifying the Trained DNN Model’s Vulnerability to Misclassification
SN - 978-989-758-513-5
AU - Agarwal H.
AU - Dorociak R.
AU - Rettberg A.
PY - 2021
SP - 211
EP - 222
DO - 10.5220/0010481302110222