Data-Driven Facial Beauty Analysis: Prediction, Retrieval and Manipulation

Abstract

In this chapter, we present a generalized data-driven facial beauty analysis framework that contains three application modules, prediction, retrieval, and manipulation. Readers are able to grasp main aspects and practical functions of facial beauty analysis via this framework. The framework is also helpful for us to design and implement applicable facial beauty analysis systems. It clearly shows relationship of different modules on the basis of the data flow and hierarchical structure of a resultant system. The chapter is organized as follows. Section 12.1 first introduces the related work of facial beauty analysis, and then gives a briefly description of the framework of facial beauty analysis. Section 12.2 describes the preprocessing and feature extraction procedures of the framework. Section 12.3 gives the definition of the beauty model. In Sect. 12.4, the facial beauty prediction method is presented. In Sect. 12.5, we discuss the beauty-oriented face retrieval problem and propose the search criteria for two application scenarios. Section 12.6 presents the facial beauty manipulation algorithm. Experimental results are shown in Sect. 12.7. We finally conclude the chapter in Sect. 12.8.