Assessment of female facial beauty based on anthropometric, non-permanent and acquisition characteristics

In this work we study the interrelation between, on the one hand, subjective perception of female facial aesthetics, and on the other hand, selected objective parameters that include facial features, photo-quality, as well as non-permanent facial characteristics. This study seeks to provide insight on the role of this specific set of features in affecting the way humans perceive facial images. The approach is novel in that it jointly considers both previous results on photo quality and beauty assessment, as well as non-permanent facial characteristics and expressions. Based on 37 such objective parameters, we construct a metric that aims to quantify modifiable parameters for aesthetics enhancement, as well as tunes systems that would seek to predict the way humans perceive facial aesthetics. The proposed metric is evaluated on a face dataset, that includes images with variations in illumination, image quality, as well as age, ethnicity and expression. We show that our approach outperforms two state of the art beauty estimation metrics. In addition we apply the designed metric in three interesting datasets, where we assess beauty in images of females before and after plastic surgery, of females across time, as well as of females famous for their beauty. We conclude by giving insight towards beauty prediction.