Recognizing Over 470 Vehicle Models with 97% Accuracy Through Learning from 190,000 Images
[Asia Economy Reporter Ki Ha-young] Hyundai Capital announced on the 6th that it has launched an "Automobile Image Recognition System" utilizing artificial intelligence (AI) technology. The system is expected to improve the accuracy of evaluations by applying it to the used car collateral loan process.
The automobile image recognition system is a solution that interprets various vehicle images (photos, website screens, etc.) into data through machine learning-based algorithm technology. Through this system, vehicle images can be recognized in real time and converted into data such as brand, model, color, and license plate, while also detecting vehicle damage and photo manipulation.
During the system development, Hyundai Capital used a total of 190,000 vehicle images of cars sold domestically as training data. The AI, after this process, verifies forgery by comparing and validating the location of license plates and vehicle colors. Thanks to extensive training, it recognizes over 470 vehicle models with an accuracy exceeding 97%.
Generally, when applying for a used car collateral loan, photos of the collateral vehicle are attached along with the loan application. During the evaluation stage, the presence of the vehicle, model, and damage status are visually verified by staff. Sometimes, discrepancies are found between the vehicle applied for and the actual vehicle. This manual inspection process requires significant manpower and time.
With the establishment of this system, Hyundai Capital has automated the loan evaluation and vehicle inspection stages, enhancing accuracy and improving the process. It is also expected to fundamentally block the possibility of loan fraud and errors during the inspection process.
A Hyundai Capital official stated, "By integrating various digital technologies, we have become the first in the industry to convert vehicle exterior images into data for use in the loan evaluation process," adding, "We plan to continuously improve evaluation accuracy through ongoing learning and utilize the accumulated data as various analytical resources."
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
