AI-Enabled Camera Traps — Gabon's ANPN Real-Time Elephant and Anti-Poaching Alert Network

Gabon is roughly 80% forest-covered — the second-highest share globally — with only about 2.2 million inhabitants, leaving vast areas of the Congo Basin effectively unmonitored for illegal logging and poaching.

Detect elephants, and potentially armed individuals, in real time using solar-powered, satellite-connected camera traps running on-device AI, without requiring local network infrastructure, and alert rangers via GPS-tagged messages over the Iridium satellite network.

Gabon's national parks agency (ANPN) and the water and forests ministry developed standardised camera-trap protocols with the University of Stirling, the Dutch start-up Hack the Planet, Panthera and the Wildlife Conservation Society. In a peer-reviewed 72-day field trial (Whytock et al., Methods in Ecology and Evolution, 2023) covering five camera units in and around Lopé National Park, the system captured over 800 images, of which 217 were elephant photographs. ANPN separately built 'Mbaza AI,' an offline image-classification tool letting rangers process camera-trap photos on laptops in the field without uploading to cloud services. GPS coordinates are deliberately stripped from shared images to stop poachers exploiting location data.

The trial achieved 82% accuracy in recognising elephants among the images captured.

Independent reporting (Mongabay) tempers the promise: a researcher surveying forest elephants nationally found camera trapping too expensive despite its precision and chose DNA analysis instead, and a Wildlife Conservation Society scientist cautioned that 'technology rarely solves problems; people do' — the cameras augment, rather than replace, ranger and researcher work, and the system cannot yet reliably identify individual elephants.