Static vs. dynamic features for automatic analysis of facial attractiveness

Analysing and measuring beauty and attractiveness has become a passion since the beginning of the human existence. Providing solutions to this mystery has been the pursuit of philosophers, artists, and anthropologists for centuries. More recently, the computer science community has attempted to propose computational models for the perception and representation of beauty by cross-fertilizing technological advancements in various fields including signal processing, computer vision and machine learning. Most of the proposed studies attempt to describe facial attractiveness via a structural model of the face obtained from a static facial image. While a static image provides limited information about facial attractiveness, using a video clip that contains information about motion, gestures, and facial expressions provides a richer and more dynamic way of analysing beauty. In this work, along with static features obtained from images, dynamic features obtained from video clips are also used to evaluate facial attractiveness. Support vector machine (SVM) and random forest (RF) are utilised to create and train models of attractiveness and evaluate the features extracted. Experimental results show that combining static and dynamic features improve performance over using either of these features alone, and SVM provides the best recognition performance.