Hailu Abenet Alazar, Kassa Banchalem Chebudie, Desta Esubalew Asmare, Adugna Fikadu Berie, Salau Ayodeji Olalekan
Department of Information Technology, College of Informatics, and University of Gondar, Gondar, Ethiopia.
Department of Computer Science, College of Informatics, University of Gondar, Gondar, Ethiopia.
Sci Rep. 2025 Aug 6;15(1):28757. doi: 10.1038/s41598-025-08076-1.
This study presents an ensemble-based approach for detecting and classifying sesame diseases using deep convolutional neural networks (CNNs). Sesame is a crucial oilseed crop that faces significant challenges from various diseases, including phyllody and bacterial blight, which adversely affect crop yield and quality. The objective of this research is to develop a robust and accurate model for identifying these diseases, leveraging the strengths of three state-of-the-art CNN architectures: ResNet-50, DenseNet-121, and Xception. The proposed ensemble model integrates these individual networks to enhance classification accuracy and improve generalization across diverse datasets. A comprehensive dataset of sesame leaf images, representing healthy, phyllody, and bacterial blight conditions was utilized to train and evaluate the models. The ensemble approach achieved an impressive overall accuracy of 96.83%, demonstrating superior performance in accurately classifying the different leaf conditions. The results highlight the effectiveness of combining multiple deep learning models, which allows for the extraction of diverse feature representations and decision-making strategies. This thesis also discusses the advantages of the ensemble methodology, including improved robustness to variations in disease symptoms and enhanced adaptability to changing agricultural practices. The findings of this research have significant implications for precision agriculture. They offer a reliable tool for the early detection and classification of sesame diseases. By enabling timely interventions, this ensemble-based framework can contribute to the sustainability and productivity of sesame cultivation, ultimately supporting food security and agricultural resilience.
本研究提出了一种基于集成的方法，利用深度卷积神经网络（CNN）对芝麻病害进行检测和分类。芝麻是一种重要的油料作物，面临着包括绿变病和细菌性叶枯病等各种病害带来的重大挑战，这些病害会对作物产量和品质产生不利影响。本研究的目的是利用三种先进的CNN架构（ResNet-50、DenseNet-121和Xception）的优势，开发一个强大且准确的模型来识别这些病害。所提出的集成模型整合了这些单个网络，以提高分类准确率并改善在不同数据集上的泛化能力。利用一个包含健康、绿变病和细菌性叶枯病状况的芝麻叶图像综合数据集来训练和评估模型。该集成方法实现了令人印象深刻的96.83%的总体准确率，在准确分类不同叶况方面表现出卓越性能。结果突出了组合多个深度学习模型的有效性，这使得能够提取多样的特征表示和决策策略。本文还讨论了集成方法的优势，包括对病害症状变化具有更高的鲁棒性以及对不断变化的农业实践具有更强的适应性。本研究的结果对精准农业具有重要意义。它们为芝麻病害的早期检测和分类提供了一个可靠的工具。通过实现及时干预，这个基于集成的框架可以促进芝麻种植的可持续性和生产力，最终支持粮食安全和农业抗灾能力。
