基于龙芯AI架构平台实现多模型算法融合的实时目标检测  

Real-time Target Detection is Realized Based on Loongson 2K1000AI Platform

作　　者：吕嫄 胡富帅 刘科文 Lv Yuan;Hu Fushuai;Liu Kewen(Department of Foundational Teaching,Wuhu Institute of Technology,Wuhu,China)

机构地区：[1]芜湖职业技术学院基础教学部,安徽芜湖

出　　处：《科学技术创新》2023年第20期113-116,共4页Scientific and Technological Innovation

基　　金：安徽省教育厅项目(KJ2019A0979,2022AH052211),芜湖职业技术学院项目(Kjcxpt2020003,rc2022dtr05)。

摘　　要：针对当前目标检测算法和分类网络模型普遍存在结构复杂、参数多、不能很好的兼容各种数据等问题,本文提出一种将多个分类网络和目标检测算法融合使用的方法,在国产自主研发的龙芯架构平台上,使用DEngine SDK开发套件内的相关模型、算法在距离远近、光线强弱场景下进行实验。通过实验对比分析,该方法的检测精准度均有可观的提升,具有良好的研究和应用价值。In view of the current target detection algorithms and Classful network models generally have problems such as complex structure,multiple parameters,and poor compatibility with various data,this paper proposes a method to integrate multiple Classful network and target detection algorithms.On the domestic self-developed Godson architecture platform,relevant models and algorithms in the DEngine SDK development kit are used to experiment in the scene of distance and light intensity.Through experimental comparison and analysis,the detection accuracy of this method has been significantly improved,and it has good research and application value.

关 键 词：龙芯 目标检测 深度学习 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]