Deep Performance Analysis of YOLOv5 to YOLOv11 for Detecting Corn Leaf Diseases

Abstract

Corn leaf diseases affect agriculture worldwide which leads to low production and economic concerns. Timely detection of diseases in corn leafs could improve the growth of corn. Timely detection could also optimize resource uses, decreases overall cost and confirm high-quality of the crop.  Deep learning techniques are playing important role in detection and classification of different leaf diseases. You Only Look Once (YOLO) models also follow the concept of deep learning for accurately detection and classification of the diseases in corn leafs. In the proposed work, YOLOv5, YOLO6, YOLOv8, YOLOv9, YOLOv10 and YOLOv11 models are trained on dataset from Kaggle for detection of diseases in corn leafs. Data augmentation techniques such as flip vertical, rotation between -15 degree to +15 degree, 90% clockwise rotation, shearing ±0 degree horizontal to ±15 degree vertical are applied on dataset at the time of training.  The model is evaluated by applying the metrics: Recall, Precision, mAP@50, mAP@50-95, and time taken. To validate the experimental results authors utilized 6 versions of YOLO. YOLOv9 performed best among proposed version and it is justified by the experimental results as well.