摘要
人脸检测是人脸识别相关应用的基础。从Viola-Jones检测器到复杂的卷积神经网络检测器,人脸检测算法的性能在不断提升。特征提取是人脸检测算法的关键,根据提取方式不同可分为手工设计特征和深度学习提取特征两类。实际应用中,采用深度学习技术的人脸检测算法的性能已超过手工设计特征的算法。针对近年人脸检测技术的进展,对几种典型的深度学习人脸检测算法的特征提取、网络结构和实验结果等几方面进行研究,以期寻找为进一步提升人脸检测算法的性能提供思路。
Face detection is the first step of face recognition applications. From Viola-Jones detector to complex convolution neural network （CNN） detectors, the performance of the face detection algorithm is increasing. Feature extraction is the key part of face recognition. There are two different extraction methods, manual-design features and deep-learning features. Many applications show that the latter＇s performance has surpassed that of the former＇s. Studies several classic face detection algorithms based deep learning in terms of how to extract features, network structure and experimental results in recent years.
作者
郑成浩
杨梦龙
ZHENG Cheng-hao;YANG Meng-long(College of Computer Science, Sichuan University, Chengdu 610065;School of Aeronautics and Astronautics Sichuan University, Chengdu 610065)
基金
国家重大科学仪器设备开发专项(No.2013YQ490879)
