Improved ResNet Based Apple Leaf Diseases Identification. Issue 32 (2022)

• Record Type: Journal Article
• Authors: Ding, Ruirou; Qiao, Yongliang; Yang, Xianghai; Jiang, Honghua; Zhang, Yunqi; Huang, Ziqi; Wang, Dongwei; Liu, Huixiang
• Is Part Of: IFAC-PapersOnLine. Volume 55:Issue 32(2022)
• Page Start: 78
• Page End: 82
• Publication Date: 2022
• Subjects: ResNet18 -- ResNet-CBAM -- ResNet18-RC -- Apple Leaf Disease -- Deep Learning; Automatic control -- Periodicals; 629.805
• DOI: 10.1016/j.ifacol.2022.11.118
• Languages: English
• ISSNs: 2405-8963

Abstract

Abstract: Apple leaf diseases occur frequently and have a wide variety, which seriously affects the yield and quality of apples. The real-time and accurate identification of apple leaf diseases is the premise to control the common leaf diseases of apple trees during the whole growth cycle, to improve apple quality, and to ensure high-quality apple output. To improve the accuracy of the apple leaf disease identification model, two improved schemes integrated ResNet18 network is proposed for the classification of apple leaf diseases. According to our experiments, the classification accuracies of the ResNet with CBAM module (ResNet-CBAM) and ResNet18 with random clipping branches (ResNet18-RC) are 95.2% and 97.2% respectively, which all slightly improved the performance of ResNet18. The improved ResNet18 networks boost the accuracy of apple leaf disease classification and can provide a theoretical basis for the prevention and control of apple leaf disease.