计算机科学 ›› 2021, Vol. 48 ›› Issue (4): 197-204.doi: 10.11896/jsjkx.200600033
李繁1, 严星2, 张晓宇1
LI Fan1, YAN Xing2, ZHANG Xiao-yu1
摘要: 特征脸算法是基于脸部表征的常用人脸辨识方法之一。当训练数据量较大时,不管是训练还是测试模块都非常耗时。基于此,采用CUDA并行运算架构实现GPU加速特征脸算法。针对GPU并行运算的效果取决于硬件规格、算法本身的复杂度和可并行性,以及程序开发者使用GPU的并行化方式等因素,文中首先提出在特征脸算法训练阶段的计算平均值、zero mean、正规化特征脸等计算步骤以及测试阶段的投影到特征脸空间、计算欧几里得距离等计算步骤使用GPU优化加速;其次在相应计算步骤采用不同的并行化加速方法并做出效能评估。实验结果表明,在人脸训练数据量在320~1920的范围内,各计算步骤加速效果明显。与Intel i7-5960X相比,GTX1060显示适配器在训练模块中可达到平均约71.7倍的加速效果,在测试模块中可达到平均约34.1倍的加速效果。
中图分类号:
[1]MACIEJ B,SKURSKI A,MAREK K,et al.Applications ofRay-Casting in Medical Imaging[J].Advances in Intelligent Systems & Computing,2014,283:3-14. [2]JIN X X,DAKU B,KO S B.Improved GPU SIMD control flow efficiency via hybrid warp size mechanism[J].Microprocessors and Microsystems,2014,38(7):717-729. [3]LIN Y C,WANG C C,LIN G,et al.A Simple Method to Improve the Quality of Diffusion-Weighted Magnetic Resonance Imaging with Rapid Histologic Correlation in a Murine Model[J].Molecular Imaging,2014,13:1-8. [4]WANG W,ZENG X H,WANG F H,et al.Parallel Time-Space Processing Model Based Fast N-body Simulation[J].Journal of Frontiers of Computer Science & Technology,2011,5(11):63-69. [5]SAPNA S,ANJALI R,KAMATH S N.Performance Analysisof Parallel Implementation of PCA-based Face Recognition using OpenCL[C]//2019 4th International Conference on Recent Trends on Electronics,Information,Communication & Techno-logy(RTEICT).IEEE,2019:877-881. [6]DESHPANDE N T,RAVISHANKAR S.Face Detection andRecognition using Viola-Jones algorithm and Fusion of PCA and ANN[J].Advances in Computational Ences and Technology,2017,10(5):1173-1189. [7]BANERJEE S,SCHEIRER W,BOWYER K,et al.Fast faceimage synthesis with minimal training[C]//2019 IEEE Winter Conference on Applications of Computer Vision(WACV).IEEE,2019:2126-2136. [8]MULGREW,AMY C.A geometric approach to study the relationship between maternal and fetal characteristics and the shape of placental surfaces[J].Dissertations & Theses Gradworks,2011,30(3):425-436. [9]WOO Y,YI C,YI Y.Fast PCA-based face recognition on GPUs[C]//IEEE International Conference on Acoustics,Speech & Signal Processing.IEEE,2013:2659-2663. [10]ZHANG D,MABU S,HIRASAWA K.Robust intelligent PCA-based face recognition framework using GNP-fuzzy data mining[J].IEEJ Transactions on Electrical & Electronic Engineering,2013,8(3):253-262. [11]FENG C,YU-BO T,MIN Y.Research and Design of ParallelParticle Swarm Optimization Algorithm Based on CUDA[J].Computer Science,2014(47):280-287. [12]HWANG F N,WEI Z H,HUANG T M,et al.A parallel addi-tive Schwarz preconditioned Jacobi-Davidson algorithm for polynomial eigenvalue problems in quantum dot simulation[J].Journal of Computational Physics,2010,229(8):2932-2947. [13]ZHAO T.A Convergence Analysis of the Inexact Simplified Jacobi-Davidson Algorithm for Polynomial Eigenvalue Problems[J].Journal of Scientific Computing,2018,75(3):1207-1228. [14]AYRES D L,CUMMINGS M P,BAELE G,et al.BEAGLE 3:Improved performance,scaling,and usability for a high-perfor-mance computing library for statistical phylogenetics[J].Syste-matic Biology,2019,68(6):1052-1061. [15]ANDREW R,DINGLE N.Implementing QR Factorization Updating Algorithms on GPUs[J].Parallel Computing,2014,40(7):161-172. |
[1] | 鲁晨阳, 邓苏, 马武彬, 吴亚辉, 周浩浩. 基于分层抽样优化的面向异构客户端的联邦学习 Federated Learning Based on Stratified Sampling Optimization for Heterogeneous Clients 计算机科学, 2022, 49(9): 183-193. https://doi.org/10.11896/jsjkx.220500263 |
[2] | 邵子灏, 杨世宇, 马国杰. 室内信息服务的基础——低成本定位技术研究综述 Foundation of Indoor Information Services:A Survey of Low-cost Localization Techniques 计算机科学, 2022, 49(9): 228-235. https://doi.org/10.11896/jsjkx.210900260 |
[3] | 张源, 康乐, 宫朝辉, 张志鸿. 基于Bi-LSTM的期货市场关联交易行为检测方法 Related Transaction Behavior Detection in Futures Market Based on Bi-LSTM 计算机科学, 2022, 49(7): 31-39. https://doi.org/10.11896/jsjkx.210400304 |
[4] | 孙刚, 伍江江, 陈浩, 李军, 徐仕远. 一种基于切比雪夫距离的隐式偏好多目标进化算法 Hidden Preference-based Multi-objective Evolutionary Algorithm Based on Chebyshev Distance 计算机科学, 2022, 49(6): 297-304. https://doi.org/10.11896/jsjkx.210500095 |
[5] | 王永, 崔源. 基于四边形最优圈内最短路径的旅行商问题割边方法 Cutting Edge Method for Traveling Salesman Problem Based on the Shortest Paths in Optimal Cycles of Quadrilaterals 计算机科学, 2022, 49(6A): 199-205. https://doi.org/10.11896/jsjkx.210400065 |
[6] | 李丹丹, 吴宇翔, 朱聪聪, 李仲康. 基于多种改进策略的改进麻雀搜索算法 Improved Sparrow Search Algorithm Based on A Variety of Improved Strategies 计算机科学, 2022, 49(6A): 217-222. https://doi.org/10.11896/jsjkx.210700032 |
[7] | 鲁晨阳, 邓苏, 马武彬, 吴亚辉, 周浩浩. 基于DBSCAN聚类的集群联邦学习方法 Clustered Federated Learning Methods Based on DBSCAN Clustering 计算机科学, 2022, 49(6A): 232-237. https://doi.org/10.11896/jsjkx.211100059 |
[8] | 胡聪, 何晓晖, 邵发明, 张艳武, 卢冠林, 王金康. 基于极大极稳定区域及SVM的交通标志检测 Traffic Sign Detection Based on MSERs and SVM 计算机科学, 2022, 49(6A): 325-330. https://doi.org/10.11896/jsjkx.210300117 |
[9] | 杨健楠, 张帆. 一种结合双注意力机制和层次网络结构的细碎农作物分类方法 Classification Method for Small Crops Combining Dual Attention Mechanisms and Hierarchical Network Structure 计算机科学, 2022, 49(6A): 353-357. https://doi.org/10.11896/jsjkx.210200169 |
[10] | 张嘉淏, 刘峰, 齐佳音. 一种基于Bottleneck Transformer的轻量级微表情识别架构 Lightweight Micro-expression Recognition Architecture Based on Bottleneck Transformer 计算机科学, 2022, 49(6A): 370-377. https://doi.org/10.11896/jsjkx.210500023 |
[11] | 王方红, 范兴刚, 杨静静, 周杰, 王德恩. 一种基于有向感知区域调整的强栅栏构建算法 Strong Barrier Construction Algorithm Based on Adjustment of Directional Sensing Area 计算机科学, 2022, 49(6A): 612-618. https://doi.org/10.11896/jsjkx.210300291 |
[12] | 田真真, 蒋维, 郑炳旭, 孟利民. 基于服务器集群的负载均衡优化调度算法 Load Balancing Optimization Scheduling Algorithm Based on Server Cluster 计算机科学, 2022, 49(6A): 639-644. https://doi.org/10.11896/jsjkx.210800071 |
[13] | 刘建美, 王洪, 马智. Shor整数分解算法的线路优化 Optimization for Shor's Integer Factorization Algorithm Circuit 计算机科学, 2022, 49(6A): 649-653. https://doi.org/10.11896/jsjkx.210600149 |
[14] | 陈博琛, 唐文兵, 黄鸿云, 丁佐华. 基于改进人工势场的未知障碍物无人机编队避障 Pop-up Obstacles Avoidance for UAV Formation Based on Improved Artificial Potential Field 计算机科学, 2022, 49(6A): 686-693. https://doi.org/10.11896/jsjkx.210500194 |
[15] | 张志龙, 史贤俊, 秦玉峰. 基于改进准深度算法的诊断策略优化方法 Diagnosis Strategy Optimization Method Based on Improved Quasi Depth Algorithm 计算机科学, 2022, 49(6A): 729-732. https://doi.org/10.11896/jsjkx.210700076 |
|