AUTHOR=Liu Bo , Fan Hongyu , Zhang Yuting , Cai Jinjin , Cheng Hong 

TITLE=Deep learning architectures for diagnosing the severity of apple frog-eye leaf spot disease in complex backgrounds

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 14 - 2023

YEAR=2024

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2023.1289497

DOI=10.3389/fpls.2023.1289497

ISSN=1664-462X

ABSTRACT=In precision agriculture, accurately diagnosing the severity of apple frog-eye leaf spot disease is essential for effective disease management. Traditional methods often rely on labor-intensive and subjective visual evaluations, which lack both efficiency and reliability. To address this challenge, particularly in complex orchard environments, we introduce a specialized deep learning architecture that comprises a two-stage multi-network model. The first stage utilizes an enhanced Pyramid Scene Parsing Network (L-DPNet) incorporating deformable convolutions, enabling robust apple leaf segmentation against complex backgrounds. The second stage employs an improved U-Net (D-UNet), optimized with bilinear upsampling and batch normalization, to achieve precise disease spot segmentation. Our proposed approach sets new benchmarks in performance, achieving a mean Intersection over Union (mIoU) of 91.27% for apple leaf and disease spots segmentation and a mean Pixel Accuracy (mPA) of 94.32%. Additionally, the method excels in classifying disease severity across five levels, attaining an overall precision of 94.81%. This work serves as an automated tool for field-based disease quantification, thereby facilitating data-driven decisions and treatment in agriculture.