In-car Damage Dirt and Stain Estimation with RGB Images

Sandra Dixe, João Leite, Sahar Azadi, Pedro Faria, José Mendes, Jaime Fonseca, João Borges

Abstract

Shared autonomous vehicles (SAV) numbers are going to increase over the next years. The absence of human driver will create a new paradigm for in-car safety. This paper addresses the problem, presenting a monitoring system capable of estimating the state of the car interior, namely the presence of damage, dirt and stains. We propose the use of Semantic Segmentation methods to perform appropriate pixel-wise classification of certain textures found in the car’s cabin as defect classes. Two methods, U-Net and DeepLabV3+, were trained and tested for different hiper-parameter and ablation scenarios, using RGB images. To be able to test and validate these approaches an In-car dataset was created, comprised by 1861 samples from 78 cars, and than splitted in 1303 train, 186 validation and 372 test RGB images. DeepLabV3+ showed promissing results, achieving an average accuracy for good, damage, stain and dirt of 77.17%, 58.60%, 65.81% and 68.82%, respectively.

Download


Paper Citation


in Harvard Style

Dixe S., Leite J., Azadi S., Faria P., Mendes J., Fonseca J. and Borges J. (2021). In-car Damage Dirt and Stain Estimation with RGB Images.In Proceedings of the 13th International Conference on Agents and Artificial Intelligence - Volume 2: ICAART, ISBN 978-989-758-484-8, pages 672-679. DOI: 10.5220/0010228006720679


in Bibtex Style

@conference{icaart21,
author={Sandra Dixe and João Leite and Sahar Azadi and Pedro Faria and José Mendes and Jaime Fonseca and João Borges},
title={In-car Damage Dirt and Stain Estimation with RGB Images},
booktitle={Proceedings of the 13th International Conference on Agents and Artificial Intelligence - Volume 2: ICAART,},
year={2021},
pages={672-679},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0010228006720679},
isbn={978-989-758-484-8},
}


in EndNote Style

TY - CONF

JO - Proceedings of the 13th International Conference on Agents and Artificial Intelligence - Volume 2: ICAART,
TI - In-car Damage Dirt and Stain Estimation with RGB Images
SN - 978-989-758-484-8
AU - Dixe S.
AU - Leite J.
AU - Azadi S.
AU - Faria P.
AU - Mendes J.
AU - Fonseca J.
AU - Borges J.
PY - 2021
SP - 672
EP - 679
DO - 10.5220/0010228006720679