Tianyu Ye, Zhaoming Wu, Shengqian Wang, Chengzhi Deng, Cong Tang
{"title":"Diagnosis method of kiwifruit foliar diseases based on improved YOLOv4-tiny","authors":"Tianyu Ye, Zhaoming Wu, Shengqian Wang, Chengzhi Deng, Cong Tang","doi":"10.1109/ICCEAI52939.2021.00058","DOIUrl":null,"url":null,"abstract":"To solve the problem of slow diagnosis speed of kiwifruit foliar surface diseases and insufficient diagnosis ability of small target diseases, a lightweight network model based on YOLOv4-Tiny is proposed. Firstly, by introducing a depthwise separable convolution at the end of the backbone network, the number of parameters is reduced while the accuracy of diagnosis is guaranteed, and the training and diagnosis speed is improved. Secondly, SPP-Net is introduced in the Neck to realize the fusion of multiple receptive fields and the aggregation of multi-scale information, thereby improving the diagnostic accuracy of the model. Lastly, the multi-feature fusion FPN model is modified to improve the diagnosis ability of small target diseases, and then improve the diagnosis accuracy. The experimental results show that our method is superior to YOLOv4-Tiny on mAP@O.5, diagnosis speed, model size and small target disease diagnosis ability.","PeriodicalId":331409,"journal":{"name":"2021 International Conference on Computer Engineering and Artificial Intelligence (ICCEAI)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2021-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 International Conference on Computer Engineering and Artificial Intelligence (ICCEAI)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICCEAI52939.2021.00058","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
To solve the problem of slow diagnosis speed of kiwifruit foliar surface diseases and insufficient diagnosis ability of small target diseases, a lightweight network model based on YOLOv4-Tiny is proposed. Firstly, by introducing a depthwise separable convolution at the end of the backbone network, the number of parameters is reduced while the accuracy of diagnosis is guaranteed, and the training and diagnosis speed is improved. Secondly, SPP-Net is introduced in the Neck to realize the fusion of multiple receptive fields and the aggregation of multi-scale information, thereby improving the diagnostic accuracy of the model. Lastly, the multi-feature fusion FPN model is modified to improve the diagnosis ability of small target diseases, and then improve the diagnosis accuracy. The experimental results show that our method is superior to YOLOv4-Tiny on mAP@O.5, diagnosis speed, model size and small target disease diagnosis ability.