Car Damage Analysis for Insurance Market Using Convolutional Neural Networks

Abstract

As the level of competition increases, image-based vehicle claim processing is gaining an important role in the insurance industry especially in handling small but more frequent insurance claims. In this study, we explore the applicability of Convolutional Neural Networks (CNNs) to determine the level of damage using damaged car images. We have used transfer learning to analyze the advantages of available object recognition models to detect and classify damage according to the damage area and the level of damage.