基于改进YOLOv5的食品包装袋缺陷检测研究

Research on Defect Detection of Food Packaging Bags Based on Improved Yolov5

罗晨昊¹

扫码查看

作者信息

1. 广东工业大学,广东广州 510006
折叠

摘要

针对传统的基于人工进行食品包装袋缺陷检测效率低下的问题,提出一种改进YOLOv5模型的食品包装袋缺陷检测方法.添加SE注意力机制,提高模型对于微小缺陷的检测能力;引入CARAFE上采样算子,利用特征图的内容信息来提高重建的质量;将激活函数替换为Mish激活函数,增强网络的准确性和泛化性.在自制的饼干包装袋数据集上,改进后的网络平均精度为88.4%,最终检测模型的mAP相较于原始模型提升了21.76%,参数量下降了9.36%.

Abstract

Aiming at the low efficiency of traditional artificial detection of food packaging bag defects,a food packaging bag defect detection method with improved YOLOv5 model was proposed.SE attention mechanism was added to improve the detection ability of the model for small defects.The CARAFE upsampling operator is introduced to improve the quality of reconstruction by using the content information of feature graph.Replace the activation function with a Mish activation function to enhance the accuracy and generalization of the network.On the self-made cookie bag data set,the average accuracy of the improved network is 88.4%,the mAP of the final detection model is increased by 21.76%,and the number of parameters is decreased by 9.36%compared with the original model.

关键词

深度学习/YOLOv5/缺陷检测/食品包装袋

Key words

deep learning/YOLOv5/defect detection/food packaging bag

引用本文复制引用

出版年

2024

绿色包装

ISSN：

段落导航