Research, “Does facial beauty exist in the eyes? A multi-method approach to interpret facial beauty predictions in machine learning models will soon be published in the Discover Artificial Intelligence Journal.
Professor Hassan Ugail, director of the Visual Computing and Intelligent Systems Centre at Bradford University, leads his research alongside the Department of Mathematics Science at Bath University, using thousands of images from diverse geographical regions and gazes. It was used in conjunction with tracking data. , when most people are asked to evaluate someone's charm, most people determine the facial features they most often see.
Researchers then trained AI models to assess which parts of the face were most prominent when it came to assess facial attractiveness.
Professor Ugeil said: Then other areas such as the lips, cheeks, and chin.
“Although existing deep learning AI models have made great strides in accurately predicting facial attractiveness, the underlying factors driving these predictions have remained largely unclear until now. Masu.
“Our approach not only increases the accuracy of AI's facial beauty prediction models, but also provides deeper insight into the role of individual facial features in predicting attractive faces.
“This has implications for areas such as cosmetic surgery, personalized beauty recommendations, and even digital media.”
This study included approximately 8,000 facial images from previously published datasets.
Applications include cosmetic surgery, digital media and advertising, and bias reduction.
Regarding methods, the team used multiple frameworks that combined three approaches. Regional attribution, salience maps are used to show the most relevant regions of the image in machine learning models. Measurement of the importance of permutation functions and the error rate of algorithms. Individual feature predictions are also a way to make machine learning results easier to understand.
Through a mix of methods and the entire dataset, researchers identified the eyes and nose as the most influential areas that determine facial attractiveness.
The team plans to further explore the relationship between cultural diversity and perceptions of facial beauty, while also investigating the application of these findings in the field of beauty assessment, including the beauty industry.
