• 

‘TwinCar (formerly NakedCar)’, an AI application developed by AgileSoDA, recognizes the damaged parts of a vehicle by classifying only the pictures that show damage and measures their severity. It is a solution that determines the type and location of damage and suggests relevant repairs. It was first used in 2018 by a Korean insurance company’s AI system to automatically estimate repair costs. TwinCar’s core algorithm’s effectiveness has been verified and introduced into the ‘AOS Alpha’ project promoted by the Korea Insurance Development Institute.

TwinCar is an AI application similar to the way a claims adjuster conducts an audit. It is designed to follow a unique process to predict the damage depth and repair cost by synthesizing the overall results based on various recognition and discrimination models to produce an automatic repair cost estimate. Since each model of TwinCar is pre-trained through various images, it is suited for image-based transfer learning. Therefore, by using images, it can be quickly applied to problems such as defect detection of PV panels, black box image analysis, facility crack safety management and facility defect prediction/ quality management.