3D Facial Ethnicity Identification Using PointNet++ with Data Augmentation Based on Farthest Point Sampling

Facial features vary among ethnic groups, and the analysis of facial shape is important in examining the similarity of these features. 3D facial point cloud data possesses diverse feature sets and has garnered significant attention for its applications in face recognition and face shape analysis. Within the domain of genetic research, the relationship between face shape and genetic factors has been particularly emphasized. Analyzing 3D facial data allows for the identification of specific genetic information that influences facial morphology. However, 3D facial data analysis via conventional image processing methods based on deep learning has proven to be challenging; thus, effective methods have not yet been established. Therefore, we propose a framework for face shape analysis by point cloud deep learning for genetic research based on PointNet and PointNet++, which allows direct input of 3D facial data, for advanced facial point cloud analysis. Furthermore, we propose a data augmentation method based on farthest point sampling that enables stable learning even with a small data set.