A lone animal silhouetted against a dusk sky, reflected in a waterhole at Selati Game Reserve

Every camera trap,
read.

Selati Game Reserve.
Limpopo, South Africa.
Proof of concept.

See the work

Camera-trap intelligence for a game reserve.

Concept proven; feasibility assessed.

The challenge

Species identification in camera-trap images was manual, time-consuming and prone to error.

What made it hard

Up to 15,000 photos and videos a month from several camera types, on a reserve where processing has to work without an internet connection.

What was built

A proof of concept for automatic species identification, with the requirements for a full monitoring platform, EcoWatch, defined with the reserve.

Where it stands.

Recorded after the July 2024 review with the reserve. Selati is named with its permission.

Concept proven; technical feasibility assessment completed in 2024.

No figure is published for this engagement, so none is shown.

Where it runs
Offline, as required
Client
Selati Game Reserve, Limpopo, South Africa
Built on
EcoWatch, our camera-trap product
A rhino in low light
A cheetah resting in grass
A spotted hyena looking at the camera
Elephants at night

Photographs on this page: Mac Stevenson, Selati Game Reserve.

What the reserve
asked for.

Selati spans about 270 square kilometres of the Limpopo Lowveld and has protected wildlife, including rhinos and cheetahs, since 1993. These are the requirements defined with the reserve’s wildlife team.

Volume
Up to 15,000 photos and videos a month from Cuddeback and Spartan cameras, with seasonal peaks.
Offline
Every step to run without an internet connection.
Species
Priority on lions, cheetahs, elephants, rhinos, leopards, hyaenas and civets.
Movement
Occupancy, heat maps and home ranges for identified species at each camera.
Alerts
Bullet wounds, snares and other threats flagged for attention, with WhatsApp and web alerts.
Measurement
Horn and tusk size estimated, with individuals above thresholds flagged.
Herds
Changes in sable and eland herd size and structure, to estimate calf survival.
Camera effort
A camera that shifts is detected, and its effort suspended until it is reset.
Reporting
Monthly reports of identified animals, and links to the tools rangers already use.
How we read images ↗Wildlife monitoring, worked through ↗

From the
reserve.

Steve Seager, Selati’s Wildlife Manager, in conversation with James Stevenson, Dscvry’s UK Director, recorded at Selati for our podcast, Let’s Talk AI.

“We’re getting so much data in that it actually takes so much time or a real specialist to come in and look at that data and pull out meaningful information.”

The data problem · Steve Seager, Wildlife Manager, Selati

“Here’s the full picture, stop just looking at it just through this keyhole.”

The full picture · Steve Seager, Wildlife Manager, Selati

“This is our findings, this is your action.”

From data to action · Steve Seager, Wildlife Manager, Selati

Listen to the full episode ↗

Worked examples of the same kind of problem: the question a team asks, and what we give back.

Which animals passed, and are they hurt?

Wildlife monitoring

Identify species, spot injuries and flag poachers from camera traps.

See how →

Is anyone where they shouldn’t be?

Perimeter breaches

Catch people and vehicles crossing the line, day or night.

See how →

Is this part defective?

Defect inspection

Catch defects on the line, not at the customer.

See how →

More real work.

Every engagement, with its reported outcome or status.

Bring us the work where wrong is expensive.

Start with the operational problem. We’ll work out the data, models and agents it needs, and where it should run.

Talk to us →
LondonDiscovery AI Limited · United Kingdom
HyderabadReDscvry Technology Private Limited · India
ChennaiReDscvry Technology Private Limited · India