深度可分卷积结合多通道注意力的垃圾图像快速分类模型

兰州理工大学学报 ›› 2023, Vol. 49 ›› Issue (3): 88-93.

• 自动化技术与计算机技术 • 上一篇下一篇

深度可分卷积结合多通道注意力的垃圾图像快速分类模型

王星^*1,*, 晏榕璟²

1.石河子开放大学, 新疆石河子 832000;
2.西交利物浦大学理学院, 江苏苏州 215123

收稿日期:2021-09-19 出版日期:2023-06-28 发布日期:2023-07-07
通讯作者: 王星(1980-),女,四川郫县人,讲师. Email:38588170@qq.com

Fast garbage images classification model based on depthwise separable convolution combined with multi-channel attention

WANG Xing¹, YAN Rong-jing²

1. Shihezi Open University, Shihezi 832000, China;
2. School of Science, Xi’an Jiaotong Liverpool University, Suzhou 215123, China;

Received:2021-09-19 Online:2023-06-28 Published:2023-07-07

摘要/Abstract

摘要： 针对传统垃圾图像分类模型结构复杂和实时性不强的问题,提出了一种深度可分卷积结合多通道注意力机制的垃圾图像快速分类模型.该模型首先利用深度卷积和逐点卷积的拼接模型构造深度可分卷积,通过减少卷积运算参数量降低模型训练时间开销;然后,引入多通道注意力机制,使模型对于强分类能力的特征具有更高的关注度;最后,在TrashNet、Garbage-classify和GINI等开源垃圾图像分类数据集上进行测试.实验结果表明,该模型相比当前主流垃圾图像分类模型,在保持识别精度较高的基础上,具有更小的时间开销和更广的检测范围.

关键词: 垃圾图像分类, 深度卷积, 逐点卷积, 多通道注意力机制

Abstract: Aiming at the problems of complex structure and poor real-time performance of traditional garbage image classification models, a fast garbage images classification model based on depthwise separable convolution combined with a multi-channel attention mechanism was proposed. Firstly, depthwise separable convolution is constructed by using deep convolution and pointwise convolution, and the training time overhead of the mode is reduced by reducing the number of convolution parameters. Then the multi-channel attention mechanism is introduced to make the model pay more attention to the features with strong classification ability. Finally, the test is performed on open-source garbage image classification data sets such as TrashNet、Garbage-classify, and GINI. The experimental results show that the model has a lower time overhead than the current mainstream garbage image classification models while maintaining a higher recognition accuracy.

Key words: garbage images classification, deep convolution, pointwise convolution, multi-channel attention

中图分类号:

TP391

王星, 晏榕璟. 深度可分卷积结合多通道注意力的垃圾图像快速分类模型[J]. 兰州理工大学学报, 2023, 49(3): 88-93.

WANG Xing, YAN Rong-jing. Fast garbage images classification model based on depthwise separable convolution combined with multi-channel attention[J]. Journal of Lanzhou University of Technology, 2023, 49(3): 88-93.

参考文献

[1] 李金玉,陈晓雷,张爱华,等.基于深度学习的垃圾分类方法综述[J].计算机工程,2022(2):1-9.
[2] ZHANG J F,LEE C.SNAP:An efficient sparse neural acceleration processor for unstructured sparse deep neural network inference[J].IEEE Journal of Solid-State Circuits,2021,56(2):636-647.
[3] KALJAHI M A.Scene image classification technique for a ubiquitous visual surveillance system[J].Multimedia Tools and Applications,2019,78(5):5791-5818.
[4] 陈智超,焦海宁,杨杰.基于改进MobileNet V2的垃圾图像分类算法[J].浙江大学学报(工学版),2021,55(8):1-10.
[5] LI Y D,GUO K,LU Y G.Cropping and attention based approach for masked face recognition[J].Applied Intelligence,2021,51(5):3012-3025.
[6] 张云,李岚.基于级联卷积神经网络的人脸特征点识别算法实现[J].兰理工大学学报,2020,46(3):105-109.
[7] 董子源,韩卫光.基于卷积神经网络的垃圾图像分类算法[J].计算机系统应用,2020,29(8):199-204.
[8] 马永杰,刘培培.基于DenseNet进化的卷积神经网络图像分类算法[J].激光与光电子学进展,2020,57(24):42-49.
[9] 朱威,屈景怡,吴仁彪.结合批归一化的直通卷积神经网络图像分类算法[J].计算机辅助设计与图形学学报,2017,29(9):1650-1657.
[10] 李鸣,张鸿.基于卷积神经网络迭代优化的图像分类算法[J].计算机工程与设计,2017,38(1):198-214.
[11] LONG F C,YAO T,QIU Z F.Coarse-to-fine localization of temporal action proposals[J].IEEE Transactions on Multimedia,2020,22(6):1577-1590.
[12] HARDING J,WALKER C.Overview of lattices of convex normal functions[J].International Journal of Intelligent Systems,2016,31(3):257-275.
[13] GAO H M,YANG Y,LI C M.Multiscale residual network with mixed depthwise convolution for hyperspectral image classification[J].IEEE Transactions on Geoscience and Remote Sensing,2021,59(4):3396-3408.
[14] QIN Y Y,CAO J T,JI X F.Fire detection method based on depthwise separable convolution and YOLOv3[J].International Journal of Automation and Computing,2021,18(2):300-310.
[15] TANG W,HUANG Y,WANG L.PokerNet:Expanding features cheaply via depthwise convolutions[J].International Journal of Automation and Computing,2021,18(3):432-442.
[16] LIANG F,TIAN Z C,DONG M.Efficient neural network using pointwise convolution kernels with linear phase constraint[J].Neurocomputing,2021,423:572-579.
[17] CHEN Y K,JUNG L,XU Y Q.FinerPCN:high fidelity point cloud completion network using pointwise convolution[J].Neurocomputing,2021,460(3):266-276.
[18] GAO M,CHEN Y H,ZHANG Z H.Classification algorithm of garbage images based on novel spatial attention mechanism and transfer learning[J].System Engineering Theory and Practice,2021,41(2):498-512.
[19] 邵江南,葛洪伟.融合残差连接与通道注意力机制的Siamese目标跟踪算法[J].计算机辅助设计与图形学学报,2021,33(2):260-269.

深度可分卷积结合多通道注意力的垃圾图像快速分类模型

Fast garbage images classification model based on depthwise separable convolution combined with multi-channel attention

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 14

编辑推荐

Metrics

本文评价

[1]	谢楠, 马振国, 唐兵, 黄煜铭, 张柯琪, 曹丹怡. 继电保护设备剩余寿命预测的智能算法研究[J]. 兰州理工大学学报, 2023, 49(2): 83-87.
[2]	苏昀暄, 涂正, 王绪安, 林川. 一种基于CKKS同态加密与神经网络的安全人脸识别方案[J]. 兰州理工大学学报, 2023, 49(2): 103-109.
[3]	王燕, 赵妮妮, 范林. 基于图神经网络特征交叉的协同过滤算法[J]. 兰州理工大学学报, 2023, 49(1): 94-102.
[4]	赵荣珍, 赵楠. 结合扰动集成RBF的故障识别方法[J]. 兰州理工大学学报, 2022, 48(6): 40-45.
[5]	冯笑, 杨雅婷, 董瑞, 艾孜麦提·艾尼瓦尔, 马博. 基于回译和集成学习的维汉神经机器翻译方法[J]. 兰州理工大学学报, 2022, 48(5): 99-106.
[6]	郭润兰, 史方青, 范雅琼, 何智. 基于卷积神经网络的壁面清洗机器人障碍物检测识别算法[J]. 兰州理工大学学报, 2022, 48(4): 83-89.
[7]	赵小强, 蒋红梅. 基于可调节判别器的领域适应[J]. 兰州理工大学学报, 2022, 48(1): 85-90.
[8]	朱昶胜,田慧星,冯文芳. 基于Adaboost算法结合DEGWO-SVM的财务困境预测[J]. 兰州理工大学学报, 2021, 47(6): 100-107.
[9]	赵荣珍, 马森财, 吴耀春. 云模型和集成极限学习机相结合的滚动轴承故障诊断方法[J]. 兰州理工大学学报, 2021, 47(4): 33-39.
[10]	王洪申, 王道俊, 曹玉法. 产品MBD数据集三维标注中工艺面的形成与校验[J]. 兰州理工大学学报, 2021, 47(3): 45-49.
[11]	王洪申, 王道俊. 产品MBD数据集三维标注的自动校验与实现[J]. 兰州理工大学学报, 2021, 47(2): 48-53.
[12]	林玉萍, 龙红, 宋盼盼, 李小棉. 由粗到精和特征筛选的精确回归预测方法及其在二语习得中的应用[J]. 兰州理工大学学报, 2020, 46(6): 104-111.
[13]	包广斌, 杨旭鹏, 康宏. 基于改进的高斯混合模型牙齿图像分割研究[J]. 兰州理工大学学报, 2020, 46(3): 100-104.
[14]	张云, 李岚. 基于级联卷积神经网络的人脸特征点识别算法实现[J]. 兰州理工大学学报, 2020, 46(3): 105-109.