基于深度学习的三维人脸识别方法研究

Study on 3D Face Recognition Method Based on Deep Learning

阐述一种基于深度学习的三维人脸识别算法。首先选取合适的人脸数据库,通过点云去噪、孔洞填充、人脸裁剪和姿态校正,生成精确的点云数据;然后,将点云数据输入至能够实现特征重用的DenseNet网络中,完成特征提取,进而采用Arcface损失函数实现人脸分类。

This paper describes a 3D face recognition algorithm based on deep learning. First, select a suitable face database, and generate accurate point cloud data through point cloud denoising, hole filling, face clipping and pose correction;Then, input the point cloud data into the Dense Net network that can realize feature reuse, complete feature extraction, and then use Arcface loss function to realize face classification.