YOLOv5s for Field Plant Disorders Detection: Challenges and Insights from a Comparative Analysis of Green bean and Faba Bean

Tarabye, Hassan H. H.; Aboelyousr, Ahmed G.; Ali, ELSAYED A. E.

doi:10.21608/agro.2025.313737.1496

YOLOv5s for Field Plant Disorders Detection: Challenges and Insights from a Comparative Analysis of Green bean and Faba Bean

Document Type : Original Article

Authors

¹ Department of Agricultural Engineering, Faculty of Agriculture, Aswan University, Egypt

² agriculture engineer - faculty of Agriculture and Natural Resources, Aswan University

³ agriculture engineering research Institute, Dokki, Giza , Egypt

10.21608/agro.2025.313737.1496

Abstract

Plant infection outbreaks are a significant concern for global food security and environmental sustainability. Green and Faba beans are crucial for providing human nutrition and animal feed. This research aims to deal with the challenges of using YOLOv5s for detecting real-time infections in bean plants in field conditions. Fourteen infections were examined in plants. Training was conducted for different infections at the individual plant level and both plants combined. The study addresses the challenges of real-time detection and recognition, intending to achieve high accuracy. The results indicated that the F1 (Harmonic Average) decreased in most Faba bean infections. This decrease was attributed to the visual similarity between Green and Faba Beans when compared to single-plant training. This suggests that the model can more accurately identify objects in images compared to the level of single-plant training in both cases. Armyworm-Beans and Armyworm-Faba-Bean show some confusion, suggesting potential feature overlap or similarity in their visual characteristics. That was logical because they are the same type of infection. While Bean common mosaic virus was often confused with Charcoal Rot and vice versa. This could be due to similar symptoms or image characteristics. Phyllocnistis citrella has a relatively high off-diagonal value with Liriomyza Sativae, indicating potential misclassification. Joint training enhances the model's ability to generalize and adapt to challenges in intricate settings, similar to how farmers adapt their practices to changing environmental conditions for the improved survey and control of the infections in the real-time, leading to more accurate and insightful decisions.

Keywords