Advanced Car Damage Assessment Using YOLOv8: A Hybrid Approach to Detection and Masking

Automatic car part segmentation and damage classification has gained considerable importance in the insurance industry for its potential to automate and streamline the damage assessment process. However, less availability of high-quality as well as publicly accessible datasets has impeded the development of robust models. In this work, we present a solution leveraging YOLOv8 technology for car part segmentation and damage type prediction. The model is trained on the Car Damage Detection dataset, which includes more than 4,000 high-resolution images Our system incorporates 21 car part segmentation classes and 8 damage type prediction classes, using semantic segmentation to identify damaged parts and estimate the extent of damage. The proposed model outperforms traditional methods in accuracy and efficiency, offering a scalable solution for automated insurance claims and accident compensation assessments.