Facial beauty assessment under unconstrained conditions

The research of facial beauty is an interdisciplinary topic involved in psychology, aesthetics, computer version and machine learning. In this paper, we propose several methods to assess facial beauty under unconstrained conditions. Our main works are as follows: First, we apply the local binary pattern (LBP) descriptor in different bins for face representation. We tried different types of LBP methods in predicting face beauty, and finally choose the method of dividing the image into 4∗4 sub-regions, and then get LBP features from those sub-regions and the whole image, respectively. Second, we operate Support Vector Machine (SVM) classifier, Random Forest, neural network and Linear Regression to do the beauty assessment task, and make a comparison of the performance of these methods. Additionally, we establish a large-scale Asian face database labeled with beauty score under unconstrained conditions.