Yield Prediction with Deep Learning on UAV Images: Banana tree application

Authors : Furkan Sönmez, Polat Ashyrov, Hayrettin Toylan

Pages : 11-22

View : 23 | Download : 78

Publication Date : 2025-12-31

Article Type : Other Papers

Abstract :Agriculture is developing with the integration of smart imaging technologies into the production, harvesting, and classification of agricultural products. This paves the way for obtaining qualified and quantitative products. The use of imaging technologies and deep learning methods in the agricultural field can increase the success of yield prediction, considering climate change and environmental conditions. This study proposes yield prediction for banana trees based on the YOLO method, using images obtained from unmanned aerial vehicles. Firstly, the performance of YOLOv8 and YOLOv9 models trained using the RoboFlow dataset is analysed. According to the comparison results, it was observed that the YOLOv9 model obtained more successful results with 87.6% mAP, 94% precision, 96% recall, and 85% F1-score. Using the YOLOv9 model, the banana yield in the trees was estimated correctly by an average of 78% in the experimental studies conducted on the images obtained by the UAV. This method provides a reliable detection approach for accurately estimating the banana tree yield but needs to be improved.
Keywords : Derin Öğrenme, İHA, YOLO, Rekolte Tahmini

ORIGINAL ARTICLE URL