Elice Group announced Wednesday that it has developed a manufacturing-specific artificial intelligence solution with Hyundai Motor's Namyang Research Center that automatically classifies, searches, and manages crash test images.
The project, led by Hyundai Motor's Namyang Research Center, was conducted over approximately eight months from late March to late November this year. The research center initiated the solution development to address the issue of test photos being stored randomly, which required significant manpower and time for photo classification and searching.
Elice Group created deep learning-based image classification and search models and built a program that combines these models according to purpose to provide an integrated service. In this process, the models were trained in a private cloud environment using Elice's modular security system, reducing the risk of data leakage and strengthening security.
The classification of crash test images, which was difficult to solve with commercial models alone, was addressed by applying a customized AI model retrained on crash test data from certified institutions. This enabled classification of approximately 60 types of test images with over 98% accuracy. Additionally, by retraining pre-trained AI with public data, the team completed a model well-suited to the Namyang Research Center's testing environment even with limited data.
The image search function allows users to select damaged areas or specific parts in photos with a mouse, then displays similar images in order based on that selection. This feature enables more precise searching of similar test cases compared to existing text and image search methods.
"Through this collaboration, we plan to continuously expand AI solutions that can be immediately applied to actual operations across various industrial sites, including manufacturing and mobility," said Kim Jae-won, CEO of Elice Group.
