The dataset collected as part of the BAMBI project combines synchronized RGB and thermal video recordings with detailed wildlife annotations and precise geo-referenced metadata. To the best of our knowledge, it is the first dataset of this scale specifically developed for UAV-based wildlife monitoring in temperate forests and adjacent habitats.
The current version includes:
- 386 paired RGB and thermal video sequences
- approximately 50 hours of UAV footage
- more than 350 GB of data
- 5,100 annotated animal tracks
- more than 1.2 million interpolated bounding boxes
- 12 species and object classes
- precise per-frame geo-referenced metadata with RTK-level accuracy
The dataset includes recordings of red deer, roe deer, and wild boar, among other species. The data were collected at various locations across Austria in forested and forest-adjacent environments and capture many of the challenges encountered in real-world UAV-based wildlife monitoring, including:
- partial or complete occlusion by tree canopies
- changing lighting conditions and shadows
- complex forest backgrounds
- seasonal variations in vegetation
- varying flight altitudes and recording conditions
- differences in animal size, visibility, and movement
By combining RGB and thermal imagery with spatial information, the dataset supports research in areas such as wildlife detection, tracking, multimodal sensor fusion, geolocation, and automated wildlife monitoring.
The BAMBI UAV Dataset is publicly available on GitHub. Further information on the dataset structure, annotations, and data access is available there.
View the dataset on GitHub: