基于二值神经网络的SSD目标检测算法  

SSD Target Detection Algorithm Based on Binary Neural Network

作　　者：申晋祥[1] 鲍美英[1] SHEN Jin-xiang;BAO Mei-ying(School of Computer and Network Engineering,Shanxi Datong University,Datong Shanxi,037009)

机构地区：[1]山西大同大学计算机与网络工程学院,山西大同037009

出　　处：《山西大同大学学报（自然科学版）》2022年第2期24-26,59,共4页Journal of Shanxi Datong University(Natural Science Edition)

基　　金：山西大同大学校级科研项目[2020YGZX016]、[2020K10];山西大同大学校级教学改革创新项目[XJG2021249]。

摘　　要：针对SSD(Single Shot MultiBox Detector)目标检测算法存在的计算量过大、占用内存空间较多而影响检测速度的问题,提出一种融合二值神经网络的改进算法。将SSD算法的浮点型的参数以及激活函数进行二值化,降低算法的时间复杂度和空间复杂度。实验结果表明,所提算法能够在基本确保目标检测准确率的情况下,减少内存占用量并有效提高检测的速度。Aiming at the problems that SSD(single shot multibox detector) target detection algorithm has too much computation and occupies much memory space, which affects the detection speed, an improved algorithm integrating binary neural network is proposed. The floating-point parameters and activation functions of SSD algorithm are binarized to reduce the time and space complexity of the algorithm. Experimental results show that the proposed algorithm can reduce the memory consumption and effectively improve the detection speed while basically ensuring the accuracy of target detection.

关 键 词：SSD算法 二值神经网络 目标检测 

分 类 号：O175.26[理学—数学]