《计算机应用》唯一官方网站 ›› 2024, Vol. 44 ›› Issue (2): 638-644.DOI: 10.11772/j.issn.1001-9081.2023030271

• 前沿与综合应用 • 上一篇    

基于轻量化YOLOv5的新型菜品识别网络

张成涵宇1, 林钰哲1, 谭程珂1, 王俊帆1,2, 顾烨婷1,2, 董哲康1,2, 高明煜1,2()   

  1. 1.杭州电子科技大学 电子信息学院,杭州 310018
    2.浙江省装备电子研究重点实验室(杭州电子科技大学),杭州 310018
  • 收稿日期:2023-03-16 修回日期:2023-04-19 接受日期:2023-04-23 发布日期:2023-05-18 出版日期:2024-02-10
  • 通讯作者: 高明煜
  • 作者简介:张成涵宇(2002—),男,浙江温州人,硕士研究生,主要研究方向:计算机视觉、网络轻量化
    林钰哲 (2002—),男,浙江金华人,硕士研究生,主要研究方向:嵌入式开发
    谭程珂(2002—),男,浙江绍兴人,硕士研究生,主要研究方向:机械臂协同控制
    王俊帆(1998—),女,浙江绍兴人,博士研究生,主要研究方向:智能交通、计算机视觉
    顾烨婷(1998—),女,浙江嘉兴人,硕士研究生,主要研究方向:计算机视觉、图像检测
    董哲康(1989—),男,浙江杭州人,副教授,博士,CCF会员,主要研究方向:神经形态计算、目标检测;
  • 基金资助:
    国家重点研发计划项目(2020YFB1710600);国家自然科学基金资助项目(62001149);浙江省重点研发计划项目(2020C01110)

New dish recognition network based on lightweight YOLOv5

Chenghanyu ZHANG1, Yuzhe LIN1, Chengke TAN1, Junfan WANG1,2, Yeting GU1,2, Zhekang DONG1,2, Mingyu GAO1,2()   

  1. 1.School of Electronics and Information Engineering,Hangzhou Dianzi University,Hangzhou Zhejiang 310018,China
    2.Zhejiang Provincial Key Laboratory of Equipment Electronics (Hangzhou Dianzi University),Hangzhou Zhejiang 310018,China
  • Received:2023-03-16 Revised:2023-04-19 Accepted:2023-04-23 Online:2023-05-18 Published:2024-02-10
  • Contact: Mingyu GAO
  • About author:ZHANG Chenghanyu, born in 2002, M. S. candidate. His research interests include computer vision, network lightweight.
    LIN Yuzhe, born in 2002, M. S. candidate. His research interests include embedded development.
    TAN Chengke, born in 2002, M. S. candidate. His research interests include robot arm cooperative control.
    WANG Junfan, born in 1998, Ph. D. candidate. Her research interests include intelligent transportation, computer vision.
    GU Yeting, born in 1998, M. S. candidate. Her research interests include computer vision, image detection.
    DONG Zhekang, born in 1989, Ph. D., associate professor. His research interests include neuromorphic computing, object detection.
  • Supported by:
    National Key Research and Development Program of China(2020YFB1710600);National Natural Science Foundation of China(62001149);Key Research and Development Program of Zhejiang Province(2020C01110)

摘要:

为了更好地满足中餐菜品识别对准确性和时效性的应用需求,设计一种新型的菜品识别网络。在原YOLOv5模型的基础上,结合Supermask方法与结构化通道剪枝对模型进行剪枝操作,并利用Int8量化技术最终实现对模型的轻量化处理,保证模型在菜品识别中兼顾准确率和速度,同时提高模型的可移植性。实验结果表明,所提模型在交并比(IoU)为0.5时,平均精度均值(mAP)达到99.00%,平均每帧识别时间达到59.54 ms,相较于原始YOLOv5降低了20 ms,且准确率基本保持一致。此外,利用Qt软件将新型菜品识别网络移植到瑞萨RZ/G2L开发板,构建智能出餐系统,可实现点餐、生成订单、自动出餐全流程,为未来真正的餐厅智能出餐系统的构建应用提供了理论与实践基础。

关键词: Supermask方法, YOLOv5, 轻量化, 网络移植, 中餐菜品识别

Abstract:

In order to better meet the accuracy and timeliness requirements of Chinese food dish recognition, a new type of dish recognition network was designed. The original YOLOv5 model was pruned by combining Supermask method and structured channel pruning method, and lightweighted finally by Int8 quantization technology. This ensured that the proposed model could balance accuracy and speed in dish recognition, achieving a good trade-off while improving the model portability. Experimental results show that the proposed model achieves a mean Average Precision (mAP) of 99.00% and an average recognition speed of 59.54 ms /frame at an Intersection over Union (IoU) of 0.5, which is 20 ms/frame faster than that of the original YOLOv5 model while maintaining the same level of accuracy. In addition, the new dish recognition network was ported to the Renesas RZ/G2L board by Qt. Based on this, an intelligent service system was constructed to realize the whole process of ordering, generating orders, and automatic meal distribution. A theoretical and practical foundation was provided for the future construction and application of truly intelligent service systems in restaurants.

Key words: Supermask method, YOLOv5, lightweight, network porting, Chinese food dish recognition

中图分类号: