C-ResNet:基于注意力机制的垃圾分类网络  

C-ResNet:Waste Classification Network Based on Attention Mechanism

作　　者：刘泽宇 方建军[1] 袁涌博 王梓鉴 LIU Zeyu;FANG Jianjun;YUAN Yongbo;WANG Zijian(College of Urban Rail Transit and Logistics,Beijing Union University,Beijing 100101,China)

机构地区：[1]北京联合大学城市轨道交通与物流学院,北京100101

出　　处：《北京联合大学学报》2025年第2期45-51,共7页Journal of Beijing Union University

摘　　要：正确的垃圾分类不仅能有效保护环境,还为后续的垃圾回收工作打下坚实基础。为了提高垃圾分类的精度和模型的泛化能力,提出了一个C-ResNet网络,采用ResNet101架构,利用CBAM注意力机制,提升对垃圾特征的提取能力,并采用EMA训练机制提高垃圾分类的效果。实验结果表明,与传统的ResNet101网络相比,C-ResNet网络的平均精度均值达到92.5%,高出原网络30个百分点,困难类别精度达到92.3%,高出原网络44.5个百分点,分类性能优于原网络。Accurate waste classification not only effectively protects the environment,but also lays a solid foundation for subsequent garbage recycling efforts.To enhance the accuracy and generalization of waste classification models,a C-ResNet network is proposed,which adopts the ResNet101 architecture,utilizes the CBAM attention mechanism to improve the extraction of waste features.Additionally,the EMA training mechanism is employed to enhance classification performance.The experimental results demonstrate that compared with the traditional ResNet101 network,the C-ResNet achieves a mean average precision of 92.5%,which is 30 percentage points higher than the original network.For challenging categories,the accuracy reaches 92.3%,exceeding the original network by 44.5 percentage points,showcasing superior classification performance.

关 键 词：垃圾分类 残差网络 CBAM注意力机制 EMA训练机制 

分 类 号：X70[环境科学与工程—环境工程] TP18[自动化与计算机技术—控制理论与控制工程] TP391[自动化与计算机技术—控制科学与工程]