一种基于改进ResNet的垃圾分类算法  

A garbage sorting algorithm based on improved ResNet

作　　者：曹文松 翟双[1] 程超[1] 张瑞婷 CAO Wensong;ZHAI Shuang;CHENG Chao;ZHANG Ruiting(School of Computer Science&Engineering,Changchun University of Technology,Changchun 130102,China)

机构地区：[1]长春工业大学计算机科学与工程学院,吉林长春130102

出　　处：《长春工业大学学报》2023年第5期416-423,共8页Journal of Changchun University of Technology

基　　金：国家自然科学基金资助项目(61903047,U20A20186);吉林省科技厅基金项目(20210201113GX);长春市科技局项目(21GD05)。

摘　　要：提出一种基于改进ResNet的垃圾分类算法,通过将通道注意力和空间注意力相结合来强化主干网络中输出的主要特征,弱化次要特征,从而提高分类精度。利用Kaggle垃圾分类数据集对所提方法进行了验证,结合评估指标及改进的损失函数,设计消融实验,对ARNet模型的识别精度和有效性进行评估和分析。实验结果表明,ARNet在Kaggle垃圾分类数据集上取得了98.16%的准确率,并且F 1-score分数达到99.18%,能够有效提高垃圾识别精度。Therefore,an improved garbage classification algorithm based on ResNet is proposed in this paper.By combining channel attention and spatial attention,the main features of the output in the trunk network are strengthened,while the secondary features are weakened,so as to improve the classification accuracy.Kaggle waste classification data set was used to verify the proposed method.Combined with evaluation indexes and improved loss functions,ablation experiments were designed to evaluate and analyze the recognition accuracy and effectiveness of ARNet model.The experimental results show that ARNet has achieved 98.16%accuracy and 99.18%F 1-score on Kaggle garbage classification data set,which can effectively improve the accuracy of garbage identification.

关 键 词：垃圾分类 注意力机制 深度学习 迁移学习 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]