WEEDS DETECTION SYSTEM: A DEEP LEARNING-BASED COMPUTER VISION PROGRAM TO DETECT WEEDS

Abstract

Weed invasion reduces agricultural productivity; timely reduction and accurate weed detection are vital to support precision farming. Traditional methods involve manual scouting and rule-based image analysis. These methods are error-prone, labor-intensive, and time-consuming. Machine Learning (ML) models are automated, but their use is limited due to generalization on complex datasets and adverse conditions. The advent of Deep Learning (DL) has transformed the field, particularly through Convolutional Neural Networks (CNNs) and object detection architectures. In this study, a robust and real-time weeds detection system is introduced utilizing the YOLOV11 object detection model. The proposed model achieved the best performance with an overall precision of 97.4%, % recall of 99.4% and a mean average Precision (mAP) of 98.4% outperforming previous versions of series and ML models. The proposed system demonstrates robustness in diverse environments and background clutter. This study shows the potential of a state-of-the-art DL model to enhance autonomous weed management in a smart agriculture system.

Keywords : Weeds, Deep Learning, Agriculture, YoloV11