基于改进YOLOv4的饮料识别算法
Beverage Identification Algorithm Based on Improved YOLOv4
沈薇 1李红梅 1陶苑 1朱学玲1
作者信息
- 1. 安徽新华学院 大数据与人工智能学院,安徽 合肥 230088
- 折叠
摘要
随着深度学习在商品识别领域的发展,饮料作为常见的商品,将饮料识别技术应用于自助饮料售卖柜中具有一定的研究意义和价值.为了减少饮料类别特征相似误检,提出了一种基于改进YOLOv4 的饮料识别算法,通过在基础网络CSPDarknet53 的每组残差模块之间增加通道注意力机制来增强饮料区域特征信息.实验结果表明,改进后的YOLOv4模型mAP值为92.43%,比改进前提高了1.74%,具有较好的实际应用价值.
Abstract
With the development of Deep Learning in the field of product identification,beverage as a common product,applying beverage recognition technology to self-service beverage cabinets has certain research significance and value.In order to reduce the misconduct of the beverage category due to similar characteristics,a beverage recognition algorithm based on improved YOLOv4 is proposed.By increasing the Channel Attention Mechanism between the residual modules of the basic network CSPDarknet53,the characteristic information of the beverage area is enhanced.The experimental results show that the mAP value of the improved YOLOv4 reaches 92.43%,which is about 1.74%higher than that before improvement,and the model has good practical application value.
关键词
饮料识别/CSPDarknet53/YOLOv4/通道注意力机制Key words
beverage identification/CSPDarknet53/YOLOv4/Channel Attention Mechanism引用本文复制引用
基金项目
安徽省省级质量工程项目(2018mooc434)
安徽省省级质量工程项目(2020jxtd120)
安徽省省级质量工程项目(2020mooc188)
安徽省大学生创新训练项目(S202212216038)
安徽省大学生创新训练项目(S202212216023)
出版年
2024
国家科技期刊平台
NSTL
万方数据