Development of a vehicle damage detection system using convolutional neural networks

Arysbek Alymkhan Bolatkhanovich, Saparkhodzhaev Nurbek Pazharbekovich, Abdrakhmanov Rustam Bakhtierovich

Images and their processing are an important part of understanding the world. Therefore, incomplete images do not allow us to determine and analyze the state of the image. This robs us of a lot of opportunities, but modern machine learning platforms can recover lost or damaged parts of such images, making it easier to understand the behind-the-scenes context and effectively analyze the created images. Extrapolation in drawing can be done by splitting local structures into unknown parts to create a single pixel (or part) of the missing part, while maintaining harmony with neighboring pixels. This deep learning app is used to identify vehicle damage and obtain initial characteristics before the event for timely insurance payments. Currently, the development of the automotive industry is directly related to the increase in the number of car accidents. Thus, insurance companies are faced with the spread of many complaints and claims at the same time. Using Mask R-CNN, owned by CNN neural networks based on machine learning and deep learning algorithms, can help solve such problems for insurance companies.

Keywords:CNN, Mask R-CNN, machine learning, Deep Learning, computer vision, neural network, Object detection, RoIAlign, RPN