YOLOv8n-Seg for Crop Weeds Detection and Instance Segmentation

The escalating global population growth has en- gendered an associated escalation in the demand for suste- nance, thereby engendering an increased reliance on agri- cultural methodologies. Presently, the agricultural sector is confronted with a dearth of manual labor due to suboptimal crop yields and inadequate precipitation. The integration of Deep Learning, Artificial Intelligence, and Machine Learning is presently being utilized to fabricate autonomous systems across diverse sectors, encompassing automotive, agriculture, and assembly line management. This study presents a novel approach for crop weeds detection utilizing a YOLOv8n-seg model with instance seg- mentation. The proposed model was trained using a combined total of 30,000 manually annotated from a gathered dataset. The plant disease detection and segmentation models proposed in this study demonstrate a mean average precision (mAP) of 99.34%, surpassing the performance of benchmark state-of- the-art models.