Car Damage Detection and Cost Prediction with MobileNetV2

In this study, an application that classifies car damages and calculates estimated repair costs has been developed with a MobileNetV2-based model. This study automates manual processes by demonstrating the applicability of deep learning techniques in the classification of vehicle damages. Users can learn the type of damage the model estimates and the corresponding cost by uploading photos of their vehicles. The dataset used in the study consists of a wide range of data covering different vehicle damage types. In this process, the MobileNetV2 model was trained with the transfer learning method and 79% accuracy rate was achieved. The application saves both time and cost for vehicle owners and insurance companies, and also minimizes human error in the detection process. It is thought that the study brings an innovative approach to vehicle damage detection systems and can inspire similar applications in other subject areas.