一种基于轻量级深度网络的目标识别方法.PDFVIP

37 3 Vol. 37 No. 3 第 卷第 期 计算机应用研究 录用定稿 Application Research of Computers Accepted Paper 一种基于轻量级深度网络的目标识别方法* 李亚辉，刘 俊 (杭州电子科技大学 通信与传输信息融合国防重点学科实验室, 杭州 310018) 摘 要：针对资源受限条件下目标识别任务，提出了一种基于轻量化深度网络的目标识别方法。通过优化卷积操作、 模型参数压缩、增强特征表达深度等网络结构的优化方法，设计并实现了针对嵌入式平台应用的轻量化网络模型结 构，使得深度网络模型在保证精度的条件下，实现了模型参数和运行所需资源的大幅缩减。实验表明，提出的轻量 化深度模型较ILSVRC-15 冠军提出的基础模型ResNet ，能够实现在ImageNet-67 数据集上，网络模型压缩为基础模 型10.2%的条件下仍保持93.5%的目标识别准确率。 关键词：深度学习；目标识别；轻量化；嵌入式应用 中图分类号：TP183 doi: 10.19734/j.issn.1001-3695.2018.07.0663 Object recognition method based on lightweight depth network Li Yahui, Liu Jun (Fundamental Science on Communication Information Transmission Fusion Technology Laboratory , Hangzhou Dianzi University, Hangzhou 310018, China) Abstract: Aiming at the task of object recognition under resource constrained condition, this paper proposed a method of object recognition based on light weight depth network. By optimizing the design method of the network structure such as convolution operation, model parameter compression and enhancement of feature expression depth , this paper designed and implemented the lightweight network model structure named Se-DResNet for embedded platform. So that the depth network model can reduce the parameters of the model and the resources needed for operation under the condition of guaranteeing the precision. The experimental results show that the lightweight depth model has better performance than that of the basic model proposed by ILSVRC-15 champion ResNet, and it can achieve the model accuracy of 93.5% under the condition that