Deep Learning based Car Damage Detection for Insurance Claim Settlement

The study presents an advanced automated damage detection system for cars based on deep learning techniques that enhance damage detection standards of the vehicle more accurately with great efficiency. The proposed system achieves automobile damage classification at a very high accuracy using pre-trained models like ResNet-50 and EfficientNet-B0. Transfer learning and ensemble learning were implemented to improve detection accuracy of the damage with the least possible number of labeled datasets. The Three-Quarter View Car Damage Dataset (TQVCD), which uses image classification based on vehicle view orientation (front and rear) and damage type (broken or crushed) classification, is included to make the damage impression comprehensive. The proposed system, thus automating the process, reduces the time for manual inspection, human errors, and optimal claim settlement for insurance companies, car rental companies, and used-car deceptive platforms. Deep learningbased approaches end up creating scalable structures that can undertake reliable performance in real time. Thus, it is very useful to be used in both the automotive and insurance sectors.