Project Name: Detection of bird nests using deep learning to support annual colonial bird monitoring efforts within Biscayne National Park
Project Link: https://maps.fiu.edu/birdnest/ — BirdNest AI: an interactive dataset dashboard, model performance metrics, and a live nest-detection demo.
Funding Agency: National Park Service (South Florida/Caribbean Inventory and Monitoring Network)
People: Levente Juhász (PI), Boyuan (Keven) Guan (Co-PI)
Update: the BirdNest AI platform has officially launched at maps.fiu.edu/birdnest, with an interactive dataset dashboard, model performance metrics, a live nest-detection demo, and a published open dataset on FIU Dataverse.
Dataset at a Glance
- 15,759 aerial images spanning 2010–2024 (14 years) across 117 survey flights
- 161,744 nest annotations across 4 occupancy classes: 138,400 occupied, 11,238 non-occupied, 8,495 chicks, 3,611 eggs
- 7 of 9 colony monitoring sites mapped within Biscayne National Park (2 sites pending GPS coordinates)
Model
- YOLOv5s6 architecture at 1280px inference resolution
- Best mAP@0.5 = 0.458 (epoch 43 of 50), trained on a 70% / 20% / 10% train / validation / test split
- Live inference demo lets users upload their own aerial photos and run detection in real time
Open Data
The full 113 GB image archive, 161,744 YOLO-format annotations, trained model weights, and training notebooks are published as a public-domain dataset (17 U.S.C. § 105) on FIU Dataverse: doi:10.34703/gzx1-9v95/UD9HTD.
Citation: Juhasz, L., & Guan, B. (2025). Colonial Bird Nest Detection Dataset: Biscayne National Park Aerial Monitoring 2010–2024. Florida International University GIS Center. Prepared for the National Park Service, South Florida/Caribbean Inventory and Monitoring Network.
Contact: Dr. Levente Juhász (PI) — ljuhasz@fiu.edu | Dr. Kevin Whelan (NPS) — kevin_r_whelan@nps.gov
Original 2023 project announcement: We are excited to announce our research project aimed at developing an innovative machine learning model for the detection of bird nests in images. These nests serve as vital indicators of aquatic ecosystem health in the Biscayne National Park. By harnessing the power of computer vision algorithms, we will create object detection models capable of identifying potential bird nests efficiently and accurately. This project is pivotal for streamlining the monitoring of bird colonies and their populations, reducing human effort, and ensuring consistency in data collection over time.
This research initiative, in collaboration with the National Park Service, will revolutionize the way we monitor and conserve our ecosystems. Together, we will pave the way for more efficient and accurate ecological assessments. #EcosystemHealth #BirdNestDetection #ConservationEfforts