Collision Analysis to Motor Dashcam Videos With YOLO and Mask R-CNN for Auto Insurance

In traditional insurance policies, the insured drivers have to go through a complicated procedure after an accident then wait a long time for human appraisers' evaluation of insurance claims. In this scenario, the insured drivers can't receive immediate financial assistance, while the financial agencies put much effort into evaluations. To address these problems, we propose a novel approach to evaluating insurance claims of accidents automatically by artificial intelligence (AI). This paper applied noted deep learning models of YOLO and Mask R-CNN to detect collisions in motor dashcam videos. This paper incorporates both of their strengths into our system. YOLO is applied to quickly detect if an accident happened. Mask R-CNN is applied to examine if a collision happened and which object was hit by the driver's motorcycle. In this paper, collision analysis was conduct with motor dashcam videos. The analysis results show our approach can effectively detect different types of collisions for further auto insurance.