Microsoft deploys AI tool to help conserve endangered giraffes

From the newsletter

Microsoft has launched a new open-source artificial intelligence tool to support the conservation of endangered wildlife in East Africa. The tool, named Generalized Image-based Re-Identification using Artificial Intelligence for Fauna Feature Extraction (GIRAFFE), identifies individual giraffes by analysing their unique spot patterns, achieving over 90 percent accuracy.

• The tool uses computer vision to automate animal identification in photographs. Previously, researchers had to manually match spot patterns to track individual giraffes, a process that was both laborious and time-consuming. 

• GIRAFFE’s open-source image-matching technology is also applicable to other species, such as zebras, leopards and whale sharks, providing low-cost, scalable solutions for data-scarce ecosystems.

More details

• The tool was developed by Microsoft’s AI for Good Lab and builds on insights from 1956, when a Canadian biologist discovered that giraffe spot patterns are unique. GIRAFFE translates this biological insight into a scalable digital solution. Researchers simply photograph the animal’s right side, upload the image and the software almost instantly matches it against a database of known individuals. GIRAFFE requires no programming skills, allowing conservationists to access it freely on GitHub and use it in the field with basic equipment. Microsoft’s Chief Data Scientist, Juan Lavista Ferres, led the tool’s development with a focus on usability and field deployment.

• The GIRAFFE project stems from over ten years of collaboration between Microsoft’s AI for Good Lab and the Wild Nature Institute. The software is already being utilised in the Masai Giraffe Conservation Project in Tanzania, where giraffe populations have declined by over 50 percent in three decades. The project aids in monitoring survival rates, movement, and reproduction. Through this partnership, conservation teams now track thousands of giraffes across East Africa.

• Data gathered using GIRAFFE helps to map migration corridors, identify high-risk zones and shape local conservation policies. The tool enhances both the speed and accuracy of population monitoring, improving response times to threats. Its open-source release eliminates the high costs typically associated with developing custom AI for conservation, potentially saving organisations tens of thousands of dollars.

• GIRAFFE uses computer vision to match individual animals based on their unique patterns. This adaptable algorithm can be retrained for other species with visual identifiers, such as zebras, tigers, leopards and whale sharks. The tool’s codebase is designed to be flexible, supporting biodiversity monitoring across various ecosystems globally. This can enable broader biodiversity monitoring throughout Africa without the need to build new systems from scratch.

• Many conservation projects in Africa lack resources for sustained data collection, limiting their ability to monitor trends, assess interventions and adapt strategies. This gap undermines conservation planning, despite the fact that long-term data is essential for tracking ecosystem changes and guiding future actions. GIRAFFE’s open-source design facilitates low-cost ecological monitoring, providing small or underfunded teams access to long-term data with minimal setup. 

Our take

• GIRAFFE’s ability to identify animals faster than any human could be a threat to the very experts it aims to support. If we treat AI as a workforce replacement, we risk sidelining rangers, ecologists and trackers whose deep field knowledge machines cannot replicate. 

• The tool should be used as a force multiplier and not a shortcut that cuts people out. AI cannot smell poachers, read fresh tracks, or earn a community’s trust. What it can do is give local experts time to focus on urgent threats. 

• But that only works if we invest in digital skills, field-ready training and infrastructure. Without that, we risk creating a sleek system that leaves the people who know the land best behind.