| Comprehensive Analysis of Deep Learning for Early and Accurate Plant Disease Identification |
| Paper ID : 1070-ICEEM2025 (R1) |
| Authors |
|
Marwa Abbas Radad1, Zienab Esam Eldin Elgohary *2, Nawal A. El-Fishawy3, Mohamed A. El-Rashidy4 1Department of Computer Science and Engineering Faculty of Electronic Engineering, Menoufia University Menofia, Egypt. 2Zainab Essam Department of Computer Science and Engineering Faculty of Electronic Engineering, Menoufia University Menofia, Egypt. 3Department of Computer Science and Engineering Faculty of Electronic Engineering, Menoufia University Menofia, Egypt. 4Department of Computer Arab East Colleges Riyadh, Kingdom of Saudi Arabia |
| Abstract |
| Agriculture plays a vital role in human life, and maintaining plant health is essential for ensuring sustainable development and global food security. In recent years, various technologies have emerged to address the challenge of early plant disease diagnosis, aiming to mitigate its impact on food quality and economic stability. Among these, deep learning has demonstrated significant promise by enabling early detection of plant diseases without the need for continuous human supervision. This study presents a comparative analysis of several state-of-the-art deep learning models, including VGG16, VGG19, XceptionNet, DenseNet201, AlexNet, ResNet50, MobileNet, and MobileNetV2, for the multi-class classification of plant leaf diseases. The performance of these models is examined under two learning strategies: transfer learning and training from scratch. The experiments are conducted using the PlantVillage dataset, which contains 54,305 annotated images spanning various plant disease categories. To evaluate the models under diverse learning scenarios, the dataset is divided into three main categories: binary-class plants, multi-class plants, and distinct-class plants. A cross-validation strategy is employed to ensure reliable and robust model comparison. Experimental results reveal that MobileNetV2 and DenseNet201 consistently achieve the highest accuracy across different testing conditions, highlighting their potential as effective solutions for real-time plant disease detection in precision agriculture. |
| Keywords |
| plant disease, deep learning, transfer learning |
| Status: Accepted |