WEED DETECTION USING YOLOV2 DEEP LEARNING ALGORITHM

Abstract

Abstract - In this paper, a deep learning algorithm called YOLOv2 (You Only Look Once version 2) is used to weed detection. The goal is to correctly recognize and distinguish between weeds and plants in photos. The method makes use of a dataset that contains photographs of plants and weeds together with matching ground truth labels, which are used as training data. The Convolutional Neural Network (CNN) layers are organized to maximize detection performance by utilizing a predetermined YOLOv2 layer graph. The goal of designing training choices is to improve the model's robustness and generalization through the use of augmentation techniques and data preprocessing. With the required data stores, layers, and programmable settings, the YOLOv2 object detector makes training and inference procedures more effective. After processing an input image through the detector, the system outputs a weed -detected image with bounding boxes surrounding instances of the recognized weeds. Metrics like training loss are used to track training progress and provide information about the model's performance and learning process. For weed detection tasks, our approach provides a scalable and efficient solution that yields significant data for agricultural applications and environmental monitoring. Keywords: We eddetection, YOLOv2, Deep learning algorithm, Plant and weed images, Ground truth labels, Data store, Convolutional Neural Network (CNN), Predefined YOLOv2 layer graph etc.