Excellence in Skin Analysis Backed By Scientific Studies
November 2, 2022
An Overview of Modiface’s Skin Analysis Tool
Modiface’s Skin Analysis tool is an AI intelligence algorithm that predicts 21 different clinical skin signs from a single selfie. These clinical skin signs include numerous different wrinkles, pigmentation related signs, vascular disorders, texture, acne, and more. From these signs, the tool can calculate 14 general skin concerns ranging from fine lines to rosacea.
This highly intelligent algorithm is the foundation of applications such as Garnier’s Skin Coach and Lancôme’s E-Youth Finder. These tools provide skincare product recommendations based on a combination of the skin signs and concerns from the algorithm and the user inputted age and skin type data. This allows brands to build a personal connection with consumers, granting them access to products specifically tailored to address their skin concerns.
A Tool Built On Expertise and Precision
The Skin analysis tool was built in collaboration between skin experts and artificial intelligence (AI) researchers. The original training dataset consisted of over 10,000 images provided with express user consent and collected in real world conditions. The selfies in this data set were collected exactly how you would take a photo when trying a skin analysis application. Next, each photo was scored on each identified clinical sign by multiple, trained dermatologists (based on L’Oréal’s expertise in skin). The average of the dermatologist scores was considered the ground truth score.
The ModiFace AI team used this data to train an artificial neural network. Ten percent of the data was reserved for testing to confirm that the neural network accurately predicts the skin signs. The result of the training and testing is an algorithm that can analyze a selfie to score the skin concerns.
Excellence In Accuracy
Our skin analysis can predict the score within 1 point in over 90% of cases, with most signs being examined on a range of 0 to 7. It can do so with variation in lighting, facial expression, ethnicity, age, and gender. When repeated in the lab, the scores show that the algorithm gets very similar accuracy to a dermatologist looking at a real individual. Bringing dermatological accuracy to consumers has been transformative for the skincare market.
A Game Changer In The Skincare Space
Modiface’s Skincare Analysis is helping users to better care for their own skin through launching with top skincare brands and retailers worldwide. In a world where personalization has become an integral part of one’s shopping experience, they see the value in leveraging a skin analysis tool that brings a trained dermatologist’s expertise directly to the consumer’s fingertips. Skincare is not one-size-fits all and knowledge is power when purchasing a product that is right for your specific skin type. Through numerous studies, detailed in published papers, we have learned key insights that give science-backed evidence to our tool’s credibility.
Looking at Over 500,000 Women With ModiFace’s Skin Analysis
With a highly accurate algorithm, we have been able to change the way we look at skin. Our Skin Analysis has helped recommend products to over 10 million users, and from this, we can learn about how skin ages and changes in a scale never before possible. We used anonymized data* of over 475,000 European women, and over 79,000 Chinese women to look for trends, which both verified expectations we had around skin aging, and helped us discover new insights.
An interesting takeaway was the finding that many skin signs increase linearly as European women age. In contrast, the skin signs were found to increase more slowly for Chinese women, often plateauing in middle age before increasing more quickly in the older age groups of Chinese women. Take a look at the full results in our paper.
* We always delete your image immediately after it is processed. We only keep anonymized information, such as your age, the region in which you used our tool, and your skin signs (which are non biometric and limited to a common numerical score). Please review our privacy policies on our brand and retail partner websites when trying a ModiFace powered Skin Analysis.
Accuracy of ModiFace’s Skin Analysis on 1,041 Diverse Women
In August of 2022, ModiFace, L'Oreal, and a third party, ECI Research, published an article in the Journal of the European Academy and Venereology. This article was a deep dive on the accuracy of our Artificial Intelligence (AI) Skin Analysis on a diverse population of women from the United States. More specifically, we looked at selfie photos of 1,041 women equally distributed across age groups, fitzpatrick skin types, and 4 self reported ethnic groups; caucasian americans, african americans, asian americans, and hispanic americans.
While we know people are more diverse than just what we looked at in this study (check out some of our other work in Europe, Japan, and on men), we wanted to take this a step further by having a third party, ECI Research, lead the implementation. They helped us recruit a more diverse pool of 50 dermatologists. From this group, each selfie was reviewed by at least 10 dermatologists.
The results were fantastic! We learned:
For 5 of the 7 reviewed skin concerns — including wrinkles, sagging, and vascular disorders — our AI's results were strongly correlated with those of the dermatologists, in many cases, out performing the dermatologists.
For the other 2 signs — cheek pores and pigmentation — these were still moderately or weakly correlated. But we seek to improve these with the launch of our Skin Analysis V2 in 2023.
Most importantly, age and ancestry has no effect on the correlation, meaning we have built a solution that works for everyone.
A Continued Journey To Improved Accuracy
One of the most useful takeaways of our third-party run test in the United States was that it helped identify areas for improvement of the algorithm and future validation studies. While Modiface’s technology reigns superior in the current market, it’s important to also identify shortcomings and areas of improvement to ensure that we maintain accuracy and excellence in our field. The research never stops as we strive to provide the world-class services that consumers deserve.