An artificial vision device (artificial eye) that assists visually impaired people has been developed with the help of artificial intelligence (AI).
A research team led by Im Mae-soon, head researcher of the Brain Convergence Research Team at the Korea Institute of Science and Technology (KIST), noted on the 20th that they have developed technology that can accurately predict the cognitive efficiency of artificial vision devices using AI.
While artificial vision devices are technologies that aid visually impaired individuals in leading everyday lives, the cognitive efficiency of these devices remains very low. Developing artificial eyes that match the level of real human vision remains a challenge in the scientific community.
One of the difficulties in developing artificial vision devices is that the development speed is very slow. These devices must undergo complex development stages, including design and production, animal testing, and clinical trials. If the expected performance is not achieved, the process requires restarting from the design phase and repeating the lengthy development process through clinical trials.
The KIST researchers developed AI technology that can reduce these complex development processes. They utilized a facial image database of 400 Koreans to train a machine learning model designed to mimic human natural face recognition abilities. They created an AI model optimized for the low-resolution black-and-white images provided by artificial vision devices.
To evaluate the performance of the AI model, the researchers measured recognition accuracy against 3,600 facial images converted to low resolution. They compared the results with cognitive experiment results of 720 facial images tested on 36 individuals with normal vision. The AI model demonstrated recognition accuracy comparable to that of real humans.
Currently, after developing the artificial vision device, cognitive experiments were conducted on able-bodied individuals, but in the future, it will be possible to perform cognitive experiment processes more quickly and efficiently using AI. This means that time and expenses dedicated to research can be significantly reduced.
Im Mae-soon said, “If the AI model developed this time is applied not only to face recognition but also to real-life image analysis, it will enable the development of a more practical artificial vision system.” She added, “We will expand the application scope of the AI-based artificial vision quality prediction model and apply it to new artificial vision technologies currently under development to expedite clinical trials and commercialization.”
References
Advanced Science(2025), DOI : https://doi.org/10.1002/advs.202405789