Deep Learning for UAV Thermal Bird Detection: Benchmarking YOLOv8–YOLOv12

This repository contains the complete dataset, trained models, and code used in the research paper “Deep Learning for UAV Thermal Bird Detection: Benchmarking YOLOv8–YOLOv12”.The study evaluates the performance of modern YOLO architectures on thermal aerial imagery collected using an unmanned aerial vehicle (UAV). The dataset includes high-resolution thermal frames containing birds at various altitudes, orientations, and environmental conditions. Contents of this deposit: dataset_all.zip (5.11 GB):Full annotated thermal dataset used for training, validation, and testing, including images, labels (YOLO format), metadata, and dataset splits. trained_models.zip (1.02 GB):Pretrained YOLOv8, YOLOv9, YOLOv10, YOLOv11, and YOLOv12 model weights used in the benchmark experiments. YOLO.ipynb:Complete Jupyter notebook including preprocessing, training, evaluation, and inference pipelines. This deposit is intended to support reproducible research in UAV-based wildlife monitoring, automated thermal detection, and deep learning for ecological applications. Please cite this dataset using the DOI generated by Zenodo.