Smart Farming Solutions: A User-Friendly GUI for Maize Tassel Estimation Using YOLO With Dynamic and Fixed Labelling, Featuring Video Support

The integration of Autonomous Aerial Vehicles (AAVs) has significantly advanced image processing and remote sensing, particularly in precision agriculture. These technologies enhance data collection and agricultural yield estimation, benefiting banks, insurance companies, and government agencies in decision-making for budget allocation and quality assessments. This study addresses the challenge of accurately quantifying corn production by developing an enhanced YOLO-v8-based deep learning model, incorporating dynamic and fixed labeling techniques, tested on 810 images and video data for real-time detection. The research utilized two primary datasets totaling 570 images. The evaluation process comprised four distinct tests: Test 1, conducted on Dataset 1 with 200 images, assessed seven attention mechanisms (SE, CBAM, GA, LKA, CA, SA, and TA) using deep learning metrics (Precision, Recall, mAP50, mAP50-95, F1-score) and statistical methods (Duncan’s test). Test 2 validated model performance on 370 images from external sources, where YOLO.SA achieved 97.48% accuracy, outperforming YOLO.LKA (95.13%). Test 3, comparing with the MTDC benchmark dataset, confirmed YOLO.SA’s accuracy at 95.93%, exceeding previous reports, while YOLO.LKA achieved 95.71%. Finally, Test 4, utilizing video-based evaluation via a developed GUI, demonstrated YOLO.SA’s superiority (95.77%) over YOLO.LKA (95.48%) and YOLO-v5 (95.72%), significantly outperforming the standard YOLO model (72.79%). This study advances computer vision in agriculture, offering a scalable, high-accuracy model for corn yield estimation, with broad applications in farming optimization, financial planning, and policy-making.