基于改进YOLOv8算法的垃圾分类识别方法研究

唐梦雨; 曾志林; 文勇

doi:10.13205/j.hjgc.202504011

基于改进YOLOv8算法的垃圾分类识别方法研究

doi: 10.13205/j.hjgc.202504011

广西民族大学人工智能学院,南宁 530006

基金项目:

广西科技基地和人才专项（桂科AD22035200）

详细信息

作者简介:
唐梦雨（1998-），女，硕士，研究方向为计算机视觉。marietang0618@foxmail.com

通讯作者:
文勇（1969-），男，硕士，高级工程师，研究方向为人工智能、计算机视觉。wenyong@gxmzu.edu.cn

计量
- 文章访问数: 213
- HTML全文浏览量: 47
- PDF下载量: 2
- 被引次数: 0
出版历程
- 收稿日期: 2024-10-15
- 录用日期: 2024-12-15
- 修回日期: 2024-11-08
- 网络出版日期: 2025-06-07
- 刊出日期: 2025-04-01

An improved garbage classification detection algorithm based on YOLOv8

School of Artificial Intelligence, Guangxi Minzu University, Nanning 530006, China

摘要

摘要: 针对当前垃圾检测算法存在对于密集目标的精确度低、易受环境背景干扰的问题，研究一种基于YOLOv8改进的垃圾检测算法YOLOv8-MA。该方法在YOLOv8网络中融合了GAM全局注力机制（global attention mechanism）和MLCA混合局部通道注意力机制（mixed local channel attention），增强对区域上下文信息的捕获能力，对输入特征图进行全局关注，实现检测精度的显著提升。此外，算法采用了EIoU损失函数，提升模型对各类样本的适应能力和模型定位精度，进一步增强整体检测性能。在TACO数据集上进行实验取得了良好效果，相比于基线模型YOLOv8，YOLOv8-MA在精确率、召回率、mAP@0.5和mAP@0.5∶0.95评价指标分别提升了1.4%、5.3%、4.2%和5.4%，表现突出，具有更好的检测效果。该研究结果证明了YOLOv8-MA算法的有效性和卓越性。此外，还通过消融实验证明了各个改进模块对提升模型性能的有效性。
- 目标检测 /
- YOLOv8 /
- 垃圾检测 /
- 注意力机制 /
- 深度学习
Abstract: To address the problems of the low accuracy for dense targets and are easily interfered by the environmental background in current garbage detection algorithms， this paper proposed a garbage detection algorithm YOLOv8-MA based on the improvement of YOLOv8. The method integrated the GAM （Global Attention Mechanism） and the MLCA （Mixed Local Channel Attention） mechanism into the YOLOv8 backbone network to enhance the capacity to capture regional context information. The GAM global attention mechanism enhanced the model's capacity to perceive and utilize global information by reducing information loss and amplifying global interactions. The MLCA mixed local channel attention mechanism focused more precisely on local feature representations， which improved the model's detection capacity to distinguish stacked garbage under complex backgrounds for achieving a significant improvement in the detection accuracy of garbage classification detection tasks， perform global attention on the input feature map and pay more attention to pixel-level information in higher-level feature maps， effectively capture the cross-dimensional interactions and establish the dependency relations between dimensions to achieve a significant improvement in detection accuracy. In addition， the algorithm adopted the EIoU loss function to enhance the adaptability of the model to adapt to various samples by optimizing the regression process of the bounding box， accelerating the convergence speed of the model for locating garbage targets， enhancing the model's detection capacity to distinguish stacked garbage under complex backgrounds. The experiments conducted on TACO datasets had achieved good results， the results showed that compared with the baseline model YOLOv8， YOLOv8-MA had improved the precision， recall， mAP@0.5 and mAP@0.5∶0.95 evaluation indicators by 1.4%， 5.3%， 4.2% and 5.4%， respectively， showing an outstanding performance and having a better detection effect. The experimental results proved the effectiveness and excellence of the YOLOv8-MA algorithm. In addition， the experiment of ablation proved the effectiveness of each improvement module in enhancing the model performance.
- object detection /
- YOLOv8 /
- garbage detection /
- attention mechanisms /
- deep learning

HTML全文

参考文献(23)

[1]	National Bureau of Statistics. China statistical yearbook 2023［M］. Beijing:China Statistics Press，2023. 国家统计局. 中国统计年鉴2023［M］. 北京:中国统计出版社，2023．
[2]	Communist Party of China Central Committee. Recommendations on formulating the 14th five-year plan for national economic and social development and the long-range objectives through the year 2035［EB/OL］.（2020-10-29）［ 2020-11-03］. https://www.gov.cn/zhengce/2020-11/03/content_5556991.htm. 中共中央关于制定国民经济和社会发展第十四个五年规划和二〇三五年远景目标的建议［EB/OL］.（2020-10-29）［ 2020-11-03］. https://www.gov.cn/zhengce/2020-11/03/content_5556991.htm.
[3]	TANG K. Research on garbage image classification method based on convolutional neural network［J］. International Journal of Computer Applications Technology Research，2020，9（1）:37.
[4]	GIRSHICK R，DONAHUE J，DARRELL T，et al. Rich feature hierarchies for accurate object detection and semantic segmentation［C］// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，Columbus，OH，USA，2014:580-587.
[5]	REDMON J，DIVVALA S，GIRSHICK R，et al. You only look once:Unified，real-time object detection［C］// Proceedings of the IEEE conference on computer vision and pattern recognition，2016:779-788.
[6]	JIN W，YU H. Cvt-assd:Convolutional vision-transformer based attentive single shot multibox detector［C］// Proceedings of the IEEE International Conference on Tools with Artificial Intelligence，2021:12364.
[7]	YAN B，FAN P，LEI X，et al. A real-time apple targets detection method for picking robot based on improved YOLOv5［C］// Proceedings of the Computer Vision and Pattern Recognition，2021，39（20）:152-160.
[8]	LI Y，HU J G，LE Y. A Waste Classification Method Based on YOLOv5 with GhostNet Integration［J］. Application of Electronic Technology，2024，50（1）:14-20. 李耀，胡军国，乐杨. 融合GhostNet的YOLOv5垃圾分类方法［J］. 电子技术应用，2024，50（1）:14-20.
[9]	DU J. A waste classification method based on YOLOv8 and transfer learning［J］. Intelligent Computing and Applications，2024，14（09）:63-69. 杜俊. 基于YOLOv8和迁移学习的垃圾分类方法［J］. 智能计算机与应用，2024，14（9）:63-69.
[10]	KUANG E Z，BHANDARI K R，GAO J. Optimizing waste management with advanced object detection for garbage classification［J］. 2024，1069:120-132.
[11]	SAWANT S，D'SOUZA L，KULKARNI A，et al. Performance evaluation of YOLO variants on marine trash images:a comparative study of YOLOv5，YOLOv7，YOLOv 8，and tiny YOLO［C］// 2024 IEEE Bangalore Humanitarian Technology Conference，2024:10564025.
[12]	HARADA R，OYAMA T，FUJIMOTO，KENJI SHIMIZU，et al. Trash detection algorithm suitable for mobile robots using improved YOLO［J］. Journal of Advanced Computatioanl Intelligence and Intelligent Informatics，2023，27（4 TN. 163）:622-631.
[13]	WANG W S，NIAN C X，ZHANG C，et al. Design of automatic waste sorting bin for non-residential areas based on YOLO v5 Model［J］. Environmental Engineering，2022，40（3）:159-165. 王文胜，年诚旭，张超，等. 基于YOLO v5模型的非住宅区自动垃圾分类箱设计［J］. 环境工程，2022，40（3）:159-165.
[14]	LE N，NGUYEN K，NGUYEN A，et al. Global-local attention for emotion recognition［J］. Neural Computing and Applications，2021，34:21625-21639.
[15]	WAN D，LU R，SHEN S，et al. Mixed local channel attention for object detection［J］. Engineering Applications of Artificial Intelligence，2023，123:106442.
[16]	ZHANG Y F，REN W，ZHANG Z，et al. Focal and efficient IOU loss for accurate bounding box regression［J］. Neurocomputing，2022，506:146-157.
[17]	HOU Q，ZHOU D，FENG J. Coordinate attention for efficient mobile network design［C］// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition，2021:13708-13717.
[18]	HU J，SHEN L，SUN G. Squeeze-and-excitation networks［C］// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition，Salt Lake City，USA，2018:7132-7141.
[19]	WOO S，PARK J，LEE J Y，et al. Cbam:convolutional block attention module［C］// Proceedings of the Europea Conference on Computer Vision，Munich，Germany，2018:3-19.
[20]	PROENA F，SIMES P. TACO:trash annotations in context for litter detection［DB］. arXiv，2020:06975.
[21]	OUYANG D，HE S，ZHANG G，et al. Efficient multi-scale attention module with cross-spatial learning［C］// Proceedings of the 2023 IEEE International Conference on Acoustics，Speech and Signal Processing，Rhodes，Greece，2023:1-5.
[22]	ZHIGANG L，BAOSHAN S，KAIYU B. Optimization of YOLOv7 based on PConv，SE attention and Wise-IoU［J］. International Journal of Computational Intelligence and Applications，2024，23（01）:2350033:1-2350033:22.
[23]	GEVORGYAN Z. SIoU loss:more powerful learning for bounding box regression［J］. Computer Engineering and Applications，2023，59（10）:187-195.