All work

Case study · Applied computer vision

Surveying 100,000 hectares of wildlife in five days.

Our in-house venture builds its own long-range drones and the AI that reads what they photograph, giving reserves a mapped, photographed record of their land instead of an estimate.

Hectares in five days
100,000
Photos in one census
1.85m
Per pixel
1.3cm

The problem

Reserves count their wildlife from the air, usually from a helicopter or light aircraft. Observers estimate what they see from a moving aircraft, and the result is a number per species and little else. There is no record of the habitat, nothing to show where each animal was, and nothing detailed enough to compare properly from one year to the next.

What we built

Veriphy is a venture incubated inside Chisl. We designed the aircraft, trained the models and built the platform, and Veriphy flies its own missions. Owning every part of it is what lets the whole job run as one system.

Our own aircraft

Long-range drones designed and built in South Africa, approved for flight beyond visual line of sight. Building our own meant we could fit the cameras the job needs and swap payloads between airframes, rather than work around the limits of an imported aircraft.

Every square metre photographed

The aircraft fly systematic transects and photograph the ground at a resolution fine enough to pick out individual animals, capturing each spot from several angles. Nothing depends on an observer spotting something at speed.

AI that reads the landscape

Computer vision models find and identify animals across millions of images. Each sighting is placed on a map and linked to its photographs, on a platform the reserve logs into. Reserves can tag what matters to them, such as a vulture nest, and the model learns to find the rest. Work is under way to extend the same models to infrastructure, water and vegetation.

Where it got to

Veriphy has counted livestock since 2022 and became fully operational for wildlife in 2026. Its first large census, run with Timbavati and Sabi Sand as a simulated annual aerial count, covered the area in five consecutive days. According to Timbavati's own write-up, the AI results aligned closely with the traditional count for larger animals and detected more grazers. The reserve was open about the limits too: small antelope are grouped together, and the models do not yet tell males from females.

The work has been covered by Daily Maverick and News24. Reserves are now asking for the same aircraft to help with anti-poaching surveillance, which is where Veriphy goes next.

What it delivered

100,000 ha
Surveyed across Timbavati and Sabi Sand in five consecutive days.
1.85M
Photographs captured and read by AI in a single census.
1.3cm
Ground resolution per pixel, with each spot photographed from several angles.

Got a problem that looks like this?

Most of what we build starts with a conversation about what is not working yet.

Start a conversation