Automated Weed Detection And Classification Using-YOLO

Authors: Mr.Rohit A. Kyarakoppa, Mr.Darshan S. Bagalkot, Mr. Manoj S. Kori, Mr. Devaraj S. Kakkali, Professor Pooja C. Shinde

Abstract: Modern agriculture practice faces increasing pressure to improve productivity while reducing chemical usage and labor dependency. One major challenge is the uncontrolled growth of agriculture weeds, which compete with crops for essential resources and significantly reduce yield. This paper presents an automated weed detection and classification system built using the YOLOv8 architecture. The model is trained on annotated crop-field images containing both crop and weed categories. The system aims to deliver real-time identification with high accuracy, enabling selective herbicide spraying and automated monitoring. The work integrates image preprocessing, model training, validation, and deployment in a practical pipeline. Experimental results demonstrate strong detection performance, reliable bounding-box prediction, and robust generalization under varied lighting and field conditions. The proposed method highlights an effective direction for precision agriculture and supports sustainable farming practices.