基于改进损失函数的残差深度学习网络人脸识别算法
首发时间:2018-01-08
摘要:近些年来,随着硬件设备的性能突破性地提高和人工智能技术的发展,对人脸识别技术研究的不断加深,利用复杂深度神经网络进行人脸识别任务被重新受到重视。本文在Caffe的深度学习框架下,对残差网络结构和损失函数进行了研究,并且提出了改进的深度学习网络算法,用以完成人脸识别的任务。通过使用CASIA WebFace人脸数据集训练网络,在具有挑战性的LFW人脸数据集上验证了提出算法的可行性和有效性,并且设计和实现了一套完整的人脸检测识别系统。
For information in English, please click here
Face Recognition Algorithm based on Improved Loss Function Residual Deep Learning Network
Abstract:In recent years, with the breakthroughs in the performance of hardware and the development of AI technology, the research of face recognition technology has been deepened. The task of face recognition using complex neural network has been emphasized. In this paper, we study the residual network structure and loss function under the framework of Caffe, and propose an improved deep learning network algorithm to complete the task of face recognition. By using CASIA WebFace face dataset to train the network, the feasibility and effectiveness of the proposed algorithm is verified on the challenging LFW face dataset, and a complete face detection and recognition system is designed and implemented.
Keywords: Pattern Recognition Face Recognition Deep Learning Residual Network
引用
No.****
动态公开评议
共计0人参与
勘误表
基于改进损失函数的残差深度学习网络人脸识别算法
评论
全部评论0/1000