Summary
This study demonstrates the application of YOLOv11 deep learning for automated sunflower head detection and yield estimation from high-resolution UAV imagery collected in sunflower fields. Using 1290 image tiles derived from 215 UAV images, the model achieved optimal performance (Precision 0.84, Recall 0.95, mAP@0.5 0.95, F1-score 0.90) when Confidence Threshold and IoU parameters were tuned to 0.50 and 0.40 respectively. The framework is proposed as a scalable, labour-saving alternative to manual yield assessment and potentially adaptable to other high-value crops.
Regional applicability
This study was conducted in Thailand and represents a precision agriculture application in a tropical/subtropical climate context. The methodology and YOLOv11-based framework are likely transferable to United Kingdom sunflower cultivation, though model retraining with UK-specific imagery and cultivars may be needed to account for differences in growing conditions, plant architecture, and image acquisition parameters.
Key measures
Precision (0.84), Recall (0.95), mAP@0.5 (0.95), F1-score (0.90); Confidence Threshold (0.50), IoU Threshold (0.40); detection accuracy and processing time reduction
Outcomes reported
The study developed and validated a YOLOv11 deep learning model for automated detection of sunflower heads in UAV imagery and estimation of crop yield. Model performance was evaluated using precision, recall, mean average precision (mAP@0.5), and F1-score metrics across optimised confidence threshold and intersection-over-union parameters.
Topic tags
Dig deeper with Pulse AI.
Ask about this record, its theme or its relevance to UK farming and policy. Pulse AI uses selected catalogue evidence and cites the sources it draws on.