The Mission

Habitat Hub, powered by Prometheus, turns aerial survey imagery into wildlife population estimates that managers can defend. African reserves base their biggest decisions on animal counts: off-takes, reintroductions, water and fire management, anti-poaching. Those counts are expensive, infrequent, and depend on a few trained observers. We make a rigorous census something a reserve can run itself. Prometheus is our geospatial engine: the methods for turning aerial and satellite data into decisions a reserve can act on. The census is the first product built on it. We start with southern African reserves, from private conservancies of around 20,000 ha to national parks of over 500,000 ha like Gonarezhou, where our co-founder already flies commercial surveys. The same core skill, counting a population you can only partly see, extends to livestock counts for lenders and insurers, forest inventory, and marine mammal surveys.

The Challenge

Censuses are costly, and how often they run depends on budget: annually at Malilangwe, twice a year in Kruger, years apart for national programmes like KAZA's US$3M 2022 census. Between counts, managers work from a snapshot. Few organisations keep the expertise in-house, yet most want to build it rather than buy a service forever. Current computer vision tools fall short in two ways: 1. Models don't travel. Detectors trained on one reserve fail on new terrain and vegetation. 2. Detection isn't a census. Animals hidden by vegetation, terrain or distance never appear in the imagery, so raw counts undercount. The hard part is turning detections into an estimate scientists and governments accept.

The solution

Habitat Hub is the analysis layer for aerial wildlife surveys. Upload drone or aircraft imagery; get species-level detections mapped to the ground and a population estimate with confidence intervals. - Detection: YOLOv11 fine-tuned on real survey imagery from reserves across Zimbabwe. Every survey we process adds new terrain to the dataset, so the model keeps getting better at the landscapes it will actually see. - Georeferencing: terrain-aware projection of oblique imagery, giving each animal a real position and distance from the flight line. - Statistical correction: distance sampling with Sentinel-2 habitat data to estimate the animals the camera missed. This is the method census programmes already accept under CITES MIKE standards. We are drone-first and priced against what reserves already pay for crewed surveys. Survey operators are our channel; reserves flying their own drones buy direct. As software, Habitat Hub gives reserves the in-house capacity they keep asking for. The census is where we start, not where we stop. Once a reserve trusts our data, the same engine serves its other monitoring needs. We are now developing fire tracking and a park-specific detection model at the request of a Mozambican park, through our operator partner.