AI-Based 'Customized Skincare Product Recommendations' Become Possible

A technology that can find and recommend cosmetics suitable for the skin based on artificial intelligence has been developed in Korea.

The Electronics and Telecommunications Research Institute (ETRI) announced on the 16th that it has developed a system that can effectively analyze the spreadability of cosmetics by measuring the texture characteristics of cosmetics using an AI-based deep learning algorithm.

Schematic comparison between expert sensory evaluation (above) and AI-based high-precision skin spreadability analysis system (below). Provided by Electronics and Telecommunications Research Institute (ETRI)

This system can identify the texture felt when cosmetics are applied to the skin with over 99% accuracy. It analyzes changes in friction measurement values, i.e., changes in spreadability, when cosmetics touch the skin through deep learning training and short-time Fourier transform (STFT) and continuous wavelet transform (CWT) techniques.

The research team reinterpreted the one-dimensional friction signals that change over time in an environment similar to applying cosmetics to the skin into a two-dimensional frequency spectrum form, extracting and analyzing mixed signals of desired time and frequency. The accuracy of the results obtained using this technique is over 99%.

The technology developed by the research team can be used to recommend the most suitable cosmetics for different age groups and genders, such as young women, middle-aged men, and infants, as well as for different seasons by analyzing the spreadability of cosmetics.

Above all, it is significant in that it can replace the existing expert sensory evaluation method, which relied on human senses to assess spreadability.

Expert sensory evaluation involves evaluators applying the product directly to the skin and scoring it subjectively based on whether it feels moist or dry, which age group might prefer it, and which season it is suitable for. This method has significant drawbacks in terms of time and cost, such as the need for expert training and supplementary tests for inconsistent results. Moreover, with the production and release of numerous products, there are clear limitations to having humans directly evaluate spreadability.

In contrast, the technology developed by the research team is expected to have high utility as it reduces errors caused by individual differences in human evaluation, provides objective assessments, and saves evaluation time and costs.

ETRI developed an AI-based cosmetics and dermatological product spreadability analysis system with a high-precision cosmetics usability testing device through joint research with TerraLeader Co., Ltd. and Amorepacific Corporation. Additionally, they received more than 10 types of formulation samples for cosmetic texture measurement from Amorepacific and conducted spreadability research with a dataset of 5,000 samples.

Yang Yong-seok, head of the Intelligent Components and Sensors Research Lab at ETRI, said, "This research is an innovative achievement that raised the level of analysis technology for cosmetics and dermatological products using deep learning models to classify commercial cosmetic creams," adding, "We hope that this technology will create new consumer trends in the K-beauty industry and lead the personalization of the beauty industry in the future."

This content was produced with the assistance of AI translation services.