基于YOLOv5的深度学习样本管理和训练自动化平台的设计与实现  

Design and Implementation of a Deep Learning Sample Management and Training Automation Platform Based on YOLOv5

作　　者：刘建丰 陈烨 焦良葆 田源 周阳 谢戴鑫 成超 Liu Jianfeng;Chen Ye;Jiao Liangbao;Tian Yuan;Zhou Yang;Xie Daixin;Cheng Chao(School of Computer Engineering,Nanjing Institute of Technology,Nanjing 211167,China)

机构地区：[1]南京工程学院计算机工程学院,南京211167

出　　处：《信息化研究》2024年第6期66-72,共7页INFORMATIZATION RESEARCH

基　　金：国家自然科学基金(No.62302211);江苏省大学生创新创业训练计划项目(No.202311276169H)。

摘　　要：随着深度学习算法的迅猛发展及其在计算机视觉和工业自动化等领域的广泛应用,数据管理的复杂性和训练过程的分散性已成为制约其实际落地的关键问题。针对这一问题,本文提出一个基于YOLOv5的深度学习样本管理和训练自动化平台。该平台集样本管理、样本标注以及训练自动化于一体,重点在于通过高效的样本管理和独特的样本标注工具简化训练过程,并通过自动化训练流程和直观的可视化界面进一步提高模型训练效率。此外,平台的模块化设计使它能够灵活适应不同用户的特定需求。在计算机视觉和工业自动化等领域,本平台展现出其独特的应用潜力。With the rapid development of deep learning algorithms and their widespread application in fields such as computer vision and industrial automation,the complexity of data management and the decentrali-zation of training processes have become key issues limiting their practical implementation.Addressing this,we propose a deep learning sample management and training automation platform based on YOLOv5.This platform integrates sample management,annotation,and training automation,focusing on simplifying the training process through efficient sample management and unique annotation tools,while enhancing model training effi-ciency through automated workflows and intuitive visualization interfaces.Moreover,its modular design allows for flexible adaptation to the specific needs of different users.The platform demonstrates its unique potential for application in areas like computer vision and industrial automation.

关 键 词：YOLOv5 PyQt5 FastDFS分布式文件系统 目标检测 软件设计 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]