Disease Detection in Tropical Tomato Leaves via Machine Learning Models

This study addresses the significant threat of tomato diseases to production in Ghana, which has led to substantial yield and quality losses, adversely affecting the livelihoods of local farmers and the availability of this essential dietary staple. Traditional disease identification methods are time-consuming and rely on subjective visual inspections, hindering early detection and control. This study develops a machine learning model capable of accurately identifying tomato plant diseases through image processing. The methodology involves processing a dataset of tomato plant images displaying healthy and diseased symptoms. The proposed model employs the YOLOv5 architecture and is deployed on a mobile platform for accessible disease identification. The model achieved a validation mAP@.5 of 0.715, demonstrating strong performance during live, on-site testing. This system provides a swift, accurate, and automated solution for detecting tomato diseases, supporting the sustainability of tomato production in Ghana.