UAV and ML-based AçaíPalm Monitoring in the Amazon: Facilitating YOLOv8 Usage in Agricultural Practices

This study explores the application of deep learning to detect and count açaípalm trees (Euterpe oleracea) in the Amazon using UAV imagery. A YOLOv8n model was trained on a georeferenced dataset collected across three municipalities at Northeast of the state of Pará, Brazil, achieving a mean average precision of 0.809 at IoU 0.5 (mAP50) and 0.312 at mAP50–95. The model was integrated into a web platform to enable non-experts, such as farmers and conservationists, to upload imagery and receive annotated results in real time. This facilitates more scalable and sustainable monitoring of açaípopulations and supports forest conservation efforts.