What makes a person beautiful? For centuries, scholars debated what comprises of beauty and how to measure beauty in a standardized, reproducible way. Some argue that beauty is not real, that it’s a myth, that the perception of beauty is learned, not developed. Others argue that perception of beauty is an “innate developmental or biological ability” (Harrar et al., 2018). Over the past few decades, scholars tried to quantify facial beauty with the advancements of computer technology and artificial intelligence.
Artificial Intelligence (AI) is defined by Merriam Webster (2018) as:
1: a branch of computer science dealing with the simulation of intelligent behavior in computers
2: the capability of a machine to imitate intelligent human behavior
One particular branch of AI — machine learning — is now being used to understand images and text. The next question becomes: Can machine learning and AI understand beauty? And more importantly, can machines make us more beautiful?
To answer these beauty questions, consider the following three topics:
AI Applications, AI Research, and AI Ideas.
AI Applications in Beauty Industry
There are many companies and organizations that already incorporate machine learning, deep learning, and AI into their beauty solutions. Here are my favourite practical AI applications in the beauty industry:
- My Beauty Matches uses machine learning and AI to “drive higher conversion rates, LTV (Life Time Value), and basket sizes” for retail partners. The machine learning model also helps partners to “discover new channels to the market and the consumer”. My Beauty Matches provides personalized recommendations helping consumers find products for their skin type. (Forbes, 2017)
- Beauty.ai created a deep learning to determine the most beautiful people on earth. The algorithm analyzed wrinkles, face symmetry, skin color, gender, age group, and ethnicity to determine the global winners. (Beauty.ai, 2016)
- Sephora uses worldwide tests and more than 1,000 combinations of foundation to help customers find their perfect match using the ColorIQ app. The app records 27 color-corrected images, eight light settings and one ultraviolet light“ to capture the skin conditions of women. (Dataconomy, 2016)
- Yahoo! Research (previously Yahoo! Labs) developed a deep learning model to categorize photographic portraits with 64 percent accuracy based on various image attributes. The study’s findings showed that “race, gender, and age are largely uncorrelated with photographic beauty.” (Predictive Analytics by Eric Siegel, 2016)
- Proven is a beauty brand that creates personalized skincare products based on the “largest beauty database in the world”. Their mission is to use artificial intelligence to improve the daily lives of women. Proven uses machine learning to learn connections between different product categories, ingredients and review ratings, then offer ingredient recommendations for consumer products. The recommendations provided by the machine learning model are “given to a cosmetic chemist who uses his expertise to create the formulations.” (Huffington Post, 2018)
- Curology is using machine learning to analyze users’ skin type, skin goals and medical history. Afterwards, users are matched with a medical professional who designs custom formulas to target individual’s skin care needs. (Huffington Post, 2018)
- Function Of Beauty (FoB) uses machine learning to create customized shampoo and conditioners. The machine learning model analyzes “hair type, hair structure, hair goals and other preferences” to come up with ingredient combinations. (Huffington Post, 2018)
- Boodles uses artificial intelligence to monitor interactions between in-store staff and online consumers in order to learn how to engage with consumers.(Crawford, 2018)
- ModiFace uses chatbot technology (Facebook Messenger), combined with Augmented Reality (AR), to help consumers choose lipstick. ModiFace uses more than 20,000 beauty products, to enable users to discover products and brands. Users can upload selfies directly into the chat to try on products virtually. Using “advanced facial tracking and simulation technology”, the consumer sees a simulation of the product on their face. Users can then purchase products leaving Facebook Messenger. (Crawford, 2018)
- Olay, a drugstore brand of Procter & Gamble, launched Skin Advisor (2017) to help women decide on beauty products. Skin Advisor is based on a deep learning algorithm which analyzes skin using selfies and recommends beauty products to purchase. This AI-powered advisor has been used “over 1.2 million times and consistently attracts 5,000 to 7,000 users every day.” (Yao, 2017)
AI Research in Beauty Industry
Artificial Intelligence research and academia is often ahead of the industry. Large budgets are dedicated to research and develop the latest machine learning, deep learning, and artificial intelligence algorithms. Reading AI research is time consuming and the process requires understanding of technical and computer science terminology. Here is my favourite AI research in the beauty industry:
- Identify face shape using machine learning and image processing techniques (Gunasinghe et al., 2016)
- Apply virtual makeup using image processing techniques (Oztel et al., 2015)
- Detect facial retouching using supervised deep learning (Bharati et al., 2016)
- Use sentiment analysis to classify cosmetic product review comments (Pugsee et al., 2017)
- Predict stacking of cosmetic ingredients for the development of alternative approaches in safety evaluation of chemicals (Nocairi et al., 2016)
AI Ideas for Beauty Industry
There are many artificial intelligence ideas to explore. AI R&D projects are greenfield projects. There are many opportunities, but coming up with ideas can be challenging. That’s why I brainstormed a few AI ideas to help you get started in the beauty industry:
- Provide women with insights related to their skin
- Show women makeup ideas by analyzing colour, style and other people’s similar facial attributes
- Help women virtually try on makeup using their own face
- Analyze makeup styles to predict social media popularity
- Learn what humans find attractive using facial analysis, analyzing facial symmetry, skin color and skin evenness (Dataconomy, 2016)
- Create better cosmetics and makeup products (Dataconomy, 2016)
- Understand human faces to predict “how a new eye shadow or face cream will actually look on the skin” (Dataconomy, 2016)
- Improve plastic surgery. “The ability to predict with near perfect accuracy what a person will look like post-surgery is vital not just for customer satisfaction but growing the entire field.” (Dataconomy, 2016)
- Improve facial reconstruction. “Rather dangerous procedures like the double-jaw surgery are increasingly common, and the opportunity to use data to keep patients safer and predict complications could prove invaluable for individuals who choose to undergo surgery.” (Dataconomy, 2016)
- Gain insight about in-store customers using visual recognition. “Retailers can gain insight in real time on what customers looked at, picked up, and didn’t buy to enrich the traditional metrics of what was bought and returned.” This AI idea can help with “inventory, visual merchandising, even shrinkage.”(Crawford, 2018)
- Understand the mood, patterns, and features of a consumer using facial recognition. Use AI and machine learning to recommend the right product and help women to apply makeup step-by-step based on the shape of their face. (Crawford, 2018)
- Predict customer orders based on age and gender using visual recognition technology (Crawford, 2018)
- Predicting return customer orders based on order history (Crawford, 2018)
- Find images and videos of women with similar facial structures using geometric transformations, triplet loss function and transfer learning in order to answer the question: “Are there other women who have a similar face to mine?” (Mira AI, 2018)
- Detect, analyze and digitally remove makeup from an “image of a human face wearing makeup” in order to predict facial beauty (Patents, 2015)
About Produvia
We help companies implement artificial intelligence technologies. Let’s chat! Visit us at produvia.com
Related Publications
Human Bias in Perception of Beauty
There are clear perceptual differences between what women consider beauty in other women and what men consider beauty in women. Men have more testosterone receptors than women, particularly in the visual regions of the brain. I assume this mechanism might be causing perceptual differences between the sexes.
Most of the beauty bias comes from datasets and algorithms. “Garbage in, garbage out” applies in this context. If the dataset is unbalanced or labelled by male researchers, the machine learning model would simply reflect the male bias. If data collection and data labelling is balanced and labelled by both sexes, then the model becomes less biased. The good news is that we can minimize human bias by what we teach computers. I previously wrote about this topic here.
Artificial Intelligence (AI) In Beauty Industry was originally published in Produvia on Medium, where people are continuing the conversation by highlighting and responding to this story.