Harnessing Deep Learning for Weed Detection in Agriculture

Abstract—Effective weed management is essential for increasing crop yields and reducing environmental harm. Traditional weed detection methods often face challenges with accuracy and efficiency, resulting in ineffective weed management. Current systems such as YOLO, while useful for real-time object detection, struggle with precise weed identification. To improve this, we propose the use of ResNet50, a deep learning model known for its robust image classification capabilities. By applying advanced preprocessing techniques, including data augmentation and noise reduction, ResNet50 enhances weed detection accuracy. Our comparative analysis reveals that ResNet50 achieves an accuracy of 98.06%, significantly outperforming YOLO, which has an accuracy of 93.95%. This advancement demonstrates ResNet50’s superior performance in weed management, leading to more efficient and sustainable agricultural practices." Keywords: Weed management, Deep learning, Image classification, ResNet50, YOLO, Accuracy, Sustainable agriculture