Leveraging Multi Modal AI Capabilities to Enhance Vehicle Insurance Claiming Process

Abstract

The cracks of the existing manual systems are evident due to the increased demand for vehicle insurance services related to vehicle accidents. The major shortcomings of previous studies have hindered the ability of industrial deployment. This paper reports a Multimodal AI approach to automate the vehicle insurance claiming process by following deep learning approaches to detect damaged parts of a vehicle, identify the damage type and severity of the damaged parts, and perform a robust and transparent claim estimate with deeper insights for stakeholders. Significant advancements were achieved in external damage detection by identifying multiple damaged parts, types and severities independently. This, combined with the multimodal approach, achieved an accuracy of 95% on final claim estimates. This enables the provision of detailed claim estimates for insurance policy holders, with maximum transparency and reasoning.