A new study published in the scientific journal Cosmetics on AI-driven dermatological diagnostics, has introduced the AI Dermatochroma Analytica (AIDA), an advanced unsupervised learning system designed to enhance skin color classification and segmentation.
Developed by researchers from Florida International University and the University of Miami Miller School of Medicine, AIDA aims to improve the precision and objectivity of skin tone analysis.
We spoke to study authors Abderrachid Hamrani, Anu Godavarty, and Robert Kirsner for their insights into the research and the potential impact of the study’s findings on the cosmetics and personal care product industries.
Transforming skin tone classification with AI
The primary goal of AIDA is to replicate how clinicians visually compare skin tones to the Fitzpatrick Skin Type (FST) reference patch while eliminating subjectivity through AI-driven clustering techniques.
“AIDA was designed to replicate how clinicians visually compare skin colors with the FST colorbar reference patch, but removes subjectivity by applying unsupervised clustering techniques,” Hamrani explained.
Unlike traditional supervised models requiring labeled datasets, AIDA’s unsupervised learning approach allows for broader adaptability across different imaging devices and lighting conditions.
The study found that AIDA achieved 97% accuracy in skin color classification, outperforming a conventional supervised convolutional neural network (CNN) model, which reached 87%.
“While AIDA has been tested under controlled conditions, ongoing work is assessing its robustness across different lighting conditions and imaging devices to ensure broader applicability,” Godavarty noted.
According to the study, “AIDA applies clustering techniques to classify skin tones without relying on labeled data, evaluating over twelve models to segment ”skin images into clusters based on color similarity, providing detailed spatial mapping aligned with dermatological standards.”
This segmentation process “reduces the uncertainty related to lighting conditions and other environmental factors, enhancing precision and consistency in skin color classification.”
Challenges and solutions in AIDA’s development
Developing AIDA was not without its challenges. “Variability in lighting conditions and imaging devices in both capturing and assessing images” posed a significant hurdle, as did “the selection of an optimal clustering algorithm for diverse skin colors” and “the need for objective ground-truth validation without relying on subjective human input,” Kirsner explained.
To address these challenges, “AIDA mimics the clinician’s visual matching process using FST reference colors but applies AI-driven clustering for consistency,” Godavarty stated. Additionally, “the LAB color space was used to minimize lighting variation effects,” but further testing will be needed to confirm the performance across more variable lighting conditions.
During the study, multiple clustering models, including K-means, hierarchical, and density-based methods, were tested, with “K-means providing the best trade-off between accuracy and efficiency,” Hamrani confirmed.
Applications in cosmetics and personal care
AIDA’s precision and adaptability open up new opportunities for the cosmetics and personal care industry. The AI system can enhance product development, improve personalized beauty solutions, and offer more precise consumer recommendations.
“AIDA can improve foundation and skin care product matching by providing precise, automated skin color classification, much like a clinician using a reference FST color patch,” Hamrani said.
The technology also has potential applications in virtual try-ons, where brands can integrate AIDA into online platforms, mobile apps, and smart mirrors “to help consumers select the best shade through real-time analysis of their skin color,” Godavarty suggested.
This advancement could lead to a more tailored and interactive shopping experience, fostering greater consumer confidence in product selections.
Future developments and industry impact
Looking ahead, the research team envisions several key advancements for AIDA. Expanding its dataset to enable finer-grained skin color classifications beyond the FST scale is a priority, and the team is also exploring applications in skin care diagnostics and theragnostics, such as identifying pigmentation disorders and predicting responses to therapy.
To support AIDA’s advancement, Kirsner concluded, “additional studies are planned to assess performance across different imaging setups and uncontrolled lighting conditions.”
The National Institutes of Health (NIH) funded the study under Grant number R01EB033413. Source: Cosmetics 2024, 11, 218. https://doi.org/10.3390/cosmetics11060218. “AI Dermatochroma Analytica (AIDA): Smart Technology for Robust Skin Color Classification and Segmentation.” Authors: Hamrani, A., et al.
