Car Damage Detection Based on Mask Scoring RCNN

Abstract

The growth of the car industry is now closely linked to the rise of the number of car accidents. As a result, insurance companies must deal with several claims at the same time while also addressing claims leakage. To resolve these issues, we propose a car damage detection system based on Mask Scoring RCNN. The experiment first makes a dataset by collecting car damage pictures of different types and on different angels for pre-processing then use Mask scoring RCNN for training. It is envisaged that this method would assist insurance in correctly classifying the damage and reducing the time spent on damage detection. The test results demonstrate that the proposed system has better masking accuracy in the case of complex images, allowing the car damage detection duty to be completed swiftly and easily.