EcoWatch at Selati Game Reserve

We built a proof of concept for camera-trap intelligence at Selati Game Reserve in Limpopo, South Africa. It identifies species automatically from camera-trap images, and it is built to run without an internet connection. The concept is proven, and a technical feasibility assessment was completed in 2024.

The problem at Selati

The reserve receives up to 15,000 photos and videos a month from several camera types. Processing has to work without an internet connection. Selati spans about 270 square kilometres of the Limpopo Lowveld and has protected wildlife, including rhinos and cheetahs, since 1993.

The requirements were defined with the reserve’s wildlife team. Offline operation is one of them, so every step has to run without an internet connection. That is what made the work hard, and it is why the offline design matters as much as the species identification itself.

What was built, and where it stands

The work is a proof of concept for automatic species identification. The requirements for a full monitoring platform, EcoWatch, were defined with the reserve. EcoWatch is our camera-trap product, and the Selati work is built on it.

The status is concept proven, and a technical feasibility assessment was completed in 2024. This page describes that proof of concept and its status, and nothing beyond it.

Wildlife monitoring more broadly

The same approach applies to our wildlife monitoring use case: camera-trap and drone imagery sorted by species, animals counted, and signs of injury flagged for rangers. The use case describes a machine at the base camp that runs the work, with images synced when the signal allows.

For the full account of the Selati work, read the Selati page.