基于辅助网络和注意力机制的皮肤病理标本检测  

Detection of Skin Pathology Specimen Based on Auxiliary Network and Attention Mechanism

作　　者：王岩[1] 崔文成[1] WANG Yan;CUI Wencheng(School of Information Science and Engineering,Shenyang University of Technology,Shenyang 110870,Liaoning)

机构地区：[1]沈阳工业大学信息科学与工程学院,辽宁沈阳110870

出　　处：《长江信息通信》2023年第7期74-77,共4页Changjiang Information & Communications

摘　　要：皮肤病理标本的批量检测对标本的送检具有重要意义,但受待检测目标较小的影响给检测任务带来巨大的挑战。针对此问题,文章提出了一种以YOLOX-DarkNet53为基本框架,结合辅助网路和注意力机制的皮肤病理标本检测模型。首先,在主干网络旁构建辅助网路的支路,以增强其特征提取能力;然后,将注意力机制应用于辅助网络和主干网络的信息融合,抑制无效信息通道,提高网络处理效率;最后,使用自建数据集对优化后的模型进行训练和测试,与原始基础网络相比,优化后的模型检测精度提高了1.26%。The batch detection of skin pathology specimens is of great significance to the submission of specimens,but the small influence of the target to be detected brings great challenges to the detection task.In response to this problem,this paper proposes a skin pathology specimen detection model based on YOLOX-DarkNet53 as the basic framework,combined with auxiliary network and attention mechanism.Firstly,a branch of the auxiliary network is built next to the backbone network to enhance its feature extraction capability;then,the attention mechanism is applied to the information fusion of the auxiliary network and the backbone network to suppress invalid information channels and improve network processing efficiency;finally,Using the selfbuilt data set to train and test the optimized model,compared with the original basic network,the detection accuracy of the optimized model increased by 1.26%.

关 键 词：皮肤病理标本 YOLOX-DarkNet53 辅助网络 注意力机制 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]