An Improved YOLOv5 for Accurate Detection and Localization of Tomato and Pepper Leaf Diseases

Abstract

Abstract Agriculture serves as a vital sector in Tunisia, supporting the nation's economy and ensuring food production. However, the detrimental impact of plant diseases on crop yield and quality presents a significant challenge for farmers. In this context, computer vision techniques have emerged as promising tools for automating disease detection processes. This paper focuses on the application of the YOLOv5 algorithm for the simultaneous detection and localization of multiple plant diseases on leaves. By using a self-generated dataset and employing techniques such as augmentation, anchor clustering, and segmentation, the study aims to enhance detection accuracy. An ablation study comparing YOLOv5s and YOLOv5x models demonstrates the superior performance of YOLOv5x, achieving a mean average precision (mAP) of 96.5%.

Code and data availability

The paper uses a self-generated dataset of tomato and pepper leaf images collected in Tunisia, but no public deposit, availability statement, or authors' URL for the dataset, annotations, code, or trained models appears in the supplied blocks. The only URL present (https://github.com/ultralytics/) is the generic YOLOv5

No evidence-backed public reproduction asset is currently recorded.