摘要
锦鲤选美的不同等级之间具有高相似度的特点,目前都是人工进行选美分级。为解决人工选美所存在的效率低、主观性强、成本高的问题,提出了一种基于迁移学习和改进ResNeXt50残差网络的锦鲤选美方法。本文首先构建了红白、大正、昭和3种锦鲤的选美等级数据集。其次,采用迁移学习策略提高训练速度,并从SE注意力模块、Hardswish激活函数和Ranger优化器3方面对ResNeXt50模型进行了改进,构建了SH-ResNeXt50锦鲤选美分级模型。试验结果表明:SH-ResNeXt50模型有效提升了锦鲤选美的等级分选能力,模型准确率达95.6%,损失值仅0.074,优于常用的AlexNet、GoogLeNet、ResNet50和ResNeXt50网络模型。最后,采用Grad-CAM分析SH-ResNeXt50模型的可解释性,结果表明SH-ResNeXt50模型和人工识别的感兴趣区域基本一致。本文所提出的方法实现了具有高相似度的锦鲤不同等级的智能分选,对其它具有高相似度的生物等级分选具有借鉴意义。
There is a high similarity among different levels of the koi for beauty pageant,and beauty grading for koi is currently done manually.To solve these problems of low efficiency,strong subjectivity and high cost of manual beauty pageants,a sorting method for koi beauty pageant was proposed based on transfer learning and improved ResNeXt50 residual network.Firstly,a rank dataset was constructed for the beauty pageant on Kohaku,Taisho and Showa koi.Secondly,the transfer learning strategy was adopted to improve the training speed and improve the ResNeXt50 model from three aspects of SE attention module,Hardswish activation function and Ranger optimizer,further a SH-ResNeXt50 classification model was proposed and constructed for koi pageant.The experimental results showed that the SH-ResNeXt50 model effectively improved the sorting ability for koi beauty pageant,with an accuracy of 95.6%and a loss value of only 0.074,which was better than the commonly used AlexNet,GoogLeNet,ResNet50 and ResNeXt50 network models.Finally,the interpretability of SH-ResNeXt50 model was analyzed by Grad-CAM,and the results showed that the regions of interest of SH-ResNeXt50 model was basically consistent with those recognized by the humans.The approach proposed realized the intelligent sorting of different levels of koi beauty pageant with high similarity,which had reference significance for other biological level sorting with high similarity.
作者
王军龙
宣魁
熊海涛
王峰
李娟
WANG Junlong;XUAN Kui;XIONG Haitao;WANG Feng;LI Juan(College of Electrical and Mechanical Engineering,Qingdao Agricultural University,Qingdao 266109,China;College of Marine Science and Engineering,Qingdao Agricultural University,Qingdao 266109,China)
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2023年第S01期330-337,共8页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金项目(32073029)
山东省自然科学基金重点项目(ZR2020KC027)
山东省研究生教育质量提升计划项目(SDYJG19134)
关键词
锦鲤选美
等级分选
迁移学习
深度学习
目标识别
目标分类
beauty pageant of koi
level sorting
transfer learning
deep learning
object detection
object classification
