This summer, Christoph Praschl represented our team at SIGGRAPH 2026 in Los Angeles, where he presented our latest work on synthetic data for wildlife detection.
The poster, “From Virtual Forests to Real Wildlife: Synthetic Non-Radiometric Thermal UAV Data for Animal Detection,” introduces a simulation pipeline that generates synthetic RGB and thermal drone imagery of forest environments, including automatically generated annotations for object detection, segmentation, and tracking.
A key focus of the work is the challenging sim-to-real transfer: training detection models on synthetic thermal data and evaluating them on real-world drone recordings of wildlife. The results show the potential of synthetic data to complement costly and difficult-to-acquire real-world datasets, particularly for applications in forest environments.
It was a great opportunity to share our research with the international computer graphics and computer vision community at SIGGRAPH Posters ’26, held July 19–23 in Los Angeles.
