三维人脸在社交媒体、数字娱乐、人脸识别、整形等领域都具有广泛的应用。前置深度摄像机在手机中的普及，使得人们可以便捷地获取自身的三维人脸。虽然很多人对自己的三维形象并不满意，但通常只有专业人员才能对得到的三维人脸进行美化。因此，如何让普通用户方便快捷地对自己的三维人脸进行美化和修饰是一个具有挑战性的问题。与现有的基于人脸美化比例规则的三维人脸美化方法不同，本项目基于人脸大数据和深度学习，通过计算机分析、整理、学习来提取人脸美学规则，并用这些规则合理美化三维人脸。具体研究内容包括适用于美化的人脸数据库的构建；基于深度学习的高精度三维人脸特征点检测、三维人脸颜值估计、三维人脸一键美化、人脸自动胖瘦调整；数据驱动的人脸肖像颜色优化；适用于整容的人脸美化、人脸衰老分析与年轻化生成等。项目最终研发出一个简单易用的智能三维人脸美化系统。普通用户、专业人士均可方便地利用该系统对三维人脸进行美化编辑。

The human face plays an important role in making a first impression and conveying emotion. As a result, beautiful faces are more pleasurable to look upon, as they are interpreted as implying purity and kindness. Scientific studies have corroborated the advantages of attractive faces. 3D face also plays an important role in social media, digital entertainment, face recognition, and plastic surgery. The popularity of front depth camera makes it possible for people to create their digital 3D faces conveniently. However, lots of people are not satisfactory with their appearances. Moreover, 3D face beautification is a skilled process which can only be accomplished by professionals. Therefore, how to enable novice users without special training to edit and retouch 3D faces intuitively and conveniently remains a challenge. Different from existing 3D face beautification methods which rely on geometric rules, this project investigates how to enhance 3D faces using 3D face databases and deep learning by learning the underlying intrinsic aesthetic rules. The research contents include 3D human face database construction for beautification, 3D face feature tracking using deep learning, 3D face attractiveness estimation, automatic 3D face beautification, automatic face reshaping, data-driven color optimization for portrait, 3D face beautification for plastic surgery, ageing process analysis, and face de-ageing. The project will develop an easy-to-use prototype 3D face beatification system. Both amateurs and professionals can use the system to edit 3D faces to make them more attractive for various important applications.