Vehicle Damage Detection and Severity Estimation

Vehicle damage assessment has become an important aspect of insurance processing and vehicle inspection process as a result of both the rapid growth in the number of vehicles being used on the roads and the rising number of accidents that involve motor vehicles. Manual inspection methods typically take longer than expected, are inconsistently applied and rely on subjective judgment of individuals, resulting in inaccurate assessments and delay in making decisions. This study presented an automated vehicle damage detection and assessment of damage severity system utilizing deep learning techniques through image processing techniques. The use of the YOLOv8 model for the detection of damaged areas of vehicles (e.g., dents, scratches, and cracks) from photos of different sides of vehicles has been implemented in the new technology to evaluate the extent of damage on vehicles. Once the damage area(s) on the vehicle have been established, the system calculates the ratio of damage and assigns a level of severity; calculates the cost of repairs based on a rules-driven approach and provides evidence of tampering if applicable; assistance in producing damage report documentation automatically. Through the evaluation of available photographs submitted with claims, the system has shown to be capable of accurately identify visible vehicle damage, classify the level of severity for damaged vehicles in a consistently reliable manner, and generate realistic estimates of repair costs. Therefore, the use of the proposed automated vehicle damage detection system has reduced the amount of time and energy spent performing damage assessments manually, thus enhancing the reliability and credibility of the results obtained from the analysis using this system. Future areas of research should look at expanding the capabilities of the proposed vehicle damage detection system to accommodate real-time analysis of video footage for continuous monitoring purposes. Further research will determine what type of advanced segmentation models (e.g., Mask R-CNN) can be used to improve the accuracy of damage feature boundary identification in determining vehicle damage severity levels, as well as using machine learning algorithms to generate accurate estimates of repair costs using historical repair data. Additional future research opportunities may include implementing the vehicle damage detection system on cloud-based platforms to improve the scalability, accuracy, and real-world applicability of the system.

[1]  Jocher, G., et al., YOLOv8: Ultralytics Real-Time Object Detection Model, Ultralytics Research and Documentation, 2023. [2]  Lin, T.-Y., Goyal, P., Girshick, R., He, K., and Dollár, P., Focal Loss for Dense Object Detection, Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017. [3]    Ma, H., and Liu, Y., Deep Learning-Based Vehicle Damage Detection Using YOLOv8, IEEE Conference on Computer Vision Applications, 2024. [4]  Mallios, D., et al., Vehicle Damage Severity Estimation for Insurance Operations Using In-the-Wild Mobile Images, IEEE Access Journal, 2023. [5]  Pérez, S. A., et al., Automated Car Damage Assessment Using Computer Vision: Insurance Company Use Case, Applied Sciences (MDPI), 2025. [6]    Qaddour, M. J., et al., Automatic Damaged Vehicle Estimator Using Enhanced Deep Learning Algorithm, Intelligent Systems with Applications, Elsevier, 2024.