Real-Time Wildlife Detection: DroneWild and Vizgard Bring AI to the Field

At the beginning of the DroneWild project, our goal was simple but ambitious :- to develop our own machine learning model capable of automatically detecting UK wildlife using drone imagery. By combining BH Wildlife Consultancy’s extensive image depository with the expertise of Peter from Addax Data Science, we’ve reached another milestone: phase two of our model, which now recognises 15 UK mammal species from aerial footage.

Some species currently have only a few hundred training annotations, while others (such as Fallow deer and Red deer) benefit from more than 250,000 annotations, thanks to the abundance of available imagery. As the project continues, we aim to expand our dataset, adding more species and strengthening existing ones with additional annotations.

From Post-Survey Analysis to Real-Time AI

Until recently, our workflow relied on a web app for analysing images and videos after surveys were completed. However, by collaborating with Vizgard, we’re now pushing the boundaries of what’s possible, exploring real-time, in-field machine learning through live drone streaming.

Our first live trial using VizEdge, a rugged and portable edge AI computer developed by Vizgard, took place at the end of September during the second DroneWild development event in Galloway Forest Park. The trial successfully demonstrated the potential of combining Vizgard’s FortifAI software platform with our own wildlife detection model, enabling automatic species identification directly in the field without the need for cloud or post-processing. During this test, the system accurately detected Red deer, Roe deer, and feral goats.

Further testing at Woburn Deer Park in Bedfordshire (part of Bedford Estates) has also shown encouraging results, identifying an even greater range of species within more complex environments.

DroneWild will continue to refine and test this integrated platform, identifying weaknesses and opportunities for improvement as we move into phase three. Machine learning remains one of the most exciting aspects of our work, and we look forward to collaborating closely with Addax AI and Vizgard over the coming months as we continue developing AI-driven wildlife monitoring.