Vehicle detection and damage identification for automated automotive dataset creation

Jack Pears; Gelayol Golcarenarenji; Alaa Mohasseb; Mohamed Bader-El-Den

Vehicle detection and damage identification for automated automotive dataset creation

Jack Pears^*
, Gelayol Golcarenarenji
, Alaa Mohasseb
, Mohamed Bader-El-Den

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Automated damage assessment in the automotive industry can provide fast, consistent results that streamline operational processes for repair and insurance organisations. However, the datasets required to train such systems are often large, time-consuming to curate, and rarely available outside proprietary sources. This study proposes a computer-vision-based preprocessing pipeline that combines RF-DETR object detection with an EfficientNetV2-M damage identification model to accelerate dataset creation and improve data quality. The vehicle detector achieved a mAP@50:95 of 94%, while the damage identification model achieved an accuracy of 97.24%.

Original language	English
Title of host publication	Proceedings of the 13th IEEE International Conference on Intelligent Systems
Publisher	Institute of Electrical and Electronics Engineers Inc.
Publication status	Accepted for publication - 31 Jul 2026
Event	The 13th IEEE International Conference on Intelligent Systems - Varna, Bulgaria Duration: 3 Sept 2026 → 5 Sept 2026 Conference number: 13 https://ieee-is-2026.blockchain2.uni-plovdiv.net/

Conference

Conference	The 13th IEEE International Conference on Intelligent Systems
Abbreviated title	IS'26
Country/Territory	Bulgaria
City	Varna
Period	3/09/26 → 5/09/26
Internet address	https://ieee-is-2026.blockchain2.uni-plovdiv.net/

Keywords

automotive damage assessment
vehicle detection
damage identification
dataset curation
deep learning
computer vision

Cite this

@inproceedings{fbce37582b0b482aa1632485260f414b,

title = "Vehicle detection and damage identification for automated automotive dataset creation",

abstract = "Automated damage assessment in the automotive industry can provide fast, consistent results that streamline operational processes for repair and insurance organisations. However, the datasets required to train such systems are often large, time-consuming to curate, and rarely available outside proprietary sources. This study proposes a computer-vision-based preprocessing pipeline that combines RF-DETR object detection with an EfficientNetV2-M damage identification model to accelerate dataset creation and improve data quality. The vehicle detector achieved a mAP@50:95 of 94\%, while the damage identification model achieved an accuracy of 97.24\%.",

keywords = "automotive damage assessment, vehicle detection, damage identification, dataset curation, deep learning, computer vision",

author = "Jack Pears and Gelayol Golcarenarenji and Alaa Mohasseb and Mohamed Bader-El-Den",

year = "2026",

month = jul,

day = "31",

language = "English",

booktitle = "Proceedings of the 13th IEEE International Conference on Intelligent Systems",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

address = "United States",

note = "The 13th IEEE International Conference on Intelligent Systems, IS'26 ; Conference date: 03-09-2026 Through 05-09-2026",

url = "https://ieee-is-2026.blockchain2.uni-plovdiv.net/",

}

Pears, J , Golcarenarenji, G , Mohasseb, A & Bader-El-Den, M 2026, Vehicle detection and damage identification for automated automotive dataset creation. in Proceedings of the 13th IEEE International Conference on Intelligent Systems. Institute of Electrical and Electronics Engineers Inc., The 13th IEEE International Conference on Intelligent Systems, Varna, Bulgaria, 3/09/26.

Vehicle detection and damage identification for automated automotive dataset creation. / Pears, Jack ; Golcarenarenji, Gelayol ; Mohasseb, Alaa et al.
Proceedings of the 13th IEEE International Conference on Intelligent Systems. Institute of Electrical and Electronics Engineers Inc., 2026.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Vehicle detection and damage identification for automated automotive dataset creation

AU - Pears, Jack

AU - Golcarenarenji, Gelayol

AU - Mohasseb, Alaa

AU - Bader-El-Den, Mohamed

N1 - Conference code: 13

PY - 2026/7/31

Y1 - 2026/7/31

N2 - Automated damage assessment in the automotive industry can provide fast, consistent results that streamline operational processes for repair and insurance organisations. However, the datasets required to train such systems are often large, time-consuming to curate, and rarely available outside proprietary sources. This study proposes a computer-vision-based preprocessing pipeline that combines RF-DETR object detection with an EfficientNetV2-M damage identification model to accelerate dataset creation and improve data quality. The vehicle detector achieved a mAP@50:95 of 94%, while the damage identification model achieved an accuracy of 97.24%.

AB - Automated damage assessment in the automotive industry can provide fast, consistent results that streamline operational processes for repair and insurance organisations. However, the datasets required to train such systems are often large, time-consuming to curate, and rarely available outside proprietary sources. This study proposes a computer-vision-based preprocessing pipeline that combines RF-DETR object detection with an EfficientNetV2-M damage identification model to accelerate dataset creation and improve data quality. The vehicle detector achieved a mAP@50:95 of 94%, while the damage identification model achieved an accuracy of 97.24%.

KW - automotive damage assessment

KW - vehicle detection

KW - damage identification

KW - dataset curation

KW - deep learning

KW - computer vision

M3 - Conference contribution

BT - Proceedings of the 13th IEEE International Conference on Intelligent Systems

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - The 13th IEEE International Conference on Intelligent Systems

Y2 - 3 September 2026 through 5 September 2026

ER -

Vehicle detection and damage identification for automated automotive dataset creation

Abstract

Conference

Keywords

Fingerprint

Cite this