It is difficult to see people beyond their appearance in the 21st century. Even the behavior of the people towards another can depend on the looks and their beauty. However, what makes a person beautiful? Are there any worldwide standards when it comes to calling a person “beautiful” or “handsome”?
Well, it may be hard for human beings to agree on general beauty standards but what if an AI does that judgment for us?[gdlr_core_space height=”30px”]
In December 2019, Turkish developer Sefik Ilkin Serengil developed an AI (using deep learning techniques) to judge people based on their looks. He worked with 60 labelers to score faces in the range of [1,5]. He used an open-source data set, from another study on AI beauty prediction, (https://github.com/HCIILAB/SCUT-FBP5500-Database-Release), consisting of 5500 images. After training and testing, AI created a list of the images with the beauty scores. When the results were sorted, the most beautiful woman became an Asian lady. These results were followed by Audrey Hepburn, Emma Watson, and Jessica Alba.
Serengil also used algorithms on women actresses in IMDb and found a list of top 100. Top 25 of the list can be found here (https://www.youtube.com/watch?v=aHzc2A5kuJM&feature=youtu.be). The more technical information about the study can be found on the developer’s blog post. (https://sefiks.com/2019/12/25/beauty-score-prediction-with-deep-learning/)[gdlr_core_space height=”30px”]
Although it’s a pleasant example of beauty perception and AI, it is not the first to work on criticizing whether a person is beautiful or not. Even some beauty contests are using AI algorithms for judging.
http://beauty.ai is the first online competition to do that, sponsored by Microsoft. The competition used 6 different judges for scoring. The judges were trained for evaluating wrinkles, face symmetry, skin color, and several other parameters. In the first year of competition, 60,000 people uploaded submitted selfies, and the most beautiful people in different age ranges were chosen.[gdlr_core_space height=”40px”]
This is exciting news since we can see what we are capable of doing with the machines and their learning abilities. However, not everything about these projects can be mentioned as angelic. The online beauty competitions that use AI judges were criticized for favoring light skin and treating light skin as a sign of beauty. It was evident that the results of the contest were mostly people with white skin, and most people didn’t find these results appropriate. The CSO (Chief Science Officer) of the beauty.ai, Alex Zhavoronkov, stated that the company didn’t build a biased algorithm. He said that if the dataset is not homogenous, your data may be biased.[gdlr_core_space height=”40px”]