基于国产软硬件的深度学习平台设计与验证  

The Design and Verification of Deep Learning Platform Based on Nationally Made Software and Hardware

作　　者：吕昊 郭江宇[1] 郝志超[1] 庄成 刘健[1] LYU Hao;GUO Jiangyu;HAO Zhichao;ZHUANG Cheng;LIU Jian(North A utomatic Control Technology Institute,Taiyuan 030006,China)

机构地区：[1]北方自动控制技术研究所,太原030006

出　　处：《火力与指挥控制》2023年第7期134-139,共6页Fire Control & Command Control

摘　　要：随着国内外政治经济形势的日益变化,实现核心技术的自主可控显得尤为重要。深度学习被应用到生活以及军事领域中,实现深度学习模型训练及推理的国产化具有重要意义。设计了一种基于全国产软硬件的深度学习平台,实现目标识别领域模型的训练,并搭建目标跟踪系统,对深度学习训练平台的实用性进行验证。With the increasing changes in the political and economic situation at home and abroad,it is particularly important to realize the independent control of core technologies.At present,deep learning is applied to life and military fields,and it is of great significance to realize the localization of deep learning training and reasoning.A deep learning platform based on nationally produced software and hardware is designed in this topic,the training of target recognition domain models is realized,and a target tracking system is built to verify the practicability of the deep learning training platform.

关 键 词：国产软硬件 深度学习训练平台 PP-YOLO 目标跟踪 

分 类 号：TP389.1[自动化与计算机技术—计算机系统结构]