Application of Yolov11 for Corn Plant Disease Detection Based on Leaf Images

Abstract

This study develops a corn leaf disease detection system using the YOLOv11 algorithm to overcome the limitations of manual identification, which is often subjective and slow. The dataset from Roboflow was converted to the object detection format with four classes (Leaf Spot, Blight, Rust, Healthy), annotated with bounding boxes, split in a 70:20:10 proportion, and optimized through preprocessing and data augmentation. The model was trained for 150 epochs, yielding an average precision of 0.785, a recall of 0.662, and an mAP@0.5 of 0.717 from 80 test images. The Healthy class performed superiorly (mAP 0.988), while the Leaf Spot class was the lowest (mAP 0.471) due to the variation of complex lesions. The confusion matrix confirmed prediction consistency. The main advantage is the detection of specific disease locations via bounding boxes, complementing previous classification approaches. This system has the potential to support automatic diagnosis and effective precision agriculture management.

Code and data availability

The paper uses a corn leaf image dataset obtained from Roboflow and a Python conversion/training script, but no public URL, deposit, or availability statement for the authors' dataset version, code, or trained model is provided. The Roboflow source is a third-party platform without a paper-specific link, and all listed

No evidence-backed public reproduction asset is currently recorded.