AMES Foundation

AI-Driven Drone Surveillance for Wildlife Protection.

TagsSustainability Vision AI

TagsSustainability Vision AI

AMES Foundation partnered with ML6 to protect Dabchick Wildlife Reserve from poaching using AI-powered drone surveillance. ML6 built a system that analyses live drone footage, detects threats such as people, vehicles or dogs, and alerts rangers within 60 seconds — helping shift conservation from reactive patrols to real-time protection.

Catch up quickly

Poaching kills hundreds of rhinos every year across South Africa. The AMES Foundation's Dabchick Wildlife Reserve spans over 1,000 hectares — far too vast for manual surveillance to cover reliably. Together with ML6, AMES deployed an AI detection system that connects directly to live drone feeds, automatically classifies threats in real time using computer vision, and sends instant WhatsApp alerts with location and timestamp to rangers on the ground. Response times dropped to under 60 seconds. The anti-poaching unit went from reactive to proactive — and the animals are better protected for it.

Continue reading

About the AMES Foundation.

The AMES Foundation gGmbH is a German non-profit on a mission to conserve wildlife, protect biodiversity, and make nature conservation financially sustainable and scalable. They work across three pillars: building a global Guardian Community of conservation partners; developing Conservation Tech including AI-driven wildlife monitoring in national parks and reserves; and directly managing species-rich Habitats — most notably the Dabchick Wildlife Reserve in South Africa, a 1,000+ hectare protected area that is home to rhinos, elands, and dozens of other threatened species.

Rangers alerted in under 60 seconds

From the moment an intruder is detected in the drone feed, security teams receive a WhatsApp alert with precise location and timestamp — in under a minute. That speed can be the difference between an interception and a los

AMES poaching notification

Vast land, finite attention
The reserve spans over 1,000 hectares. Ranger patrols and manually operated drones cannot provide constant, complete coverage across that area. Large portions of the reserve were effectively unmonitored for long stretches of time.
Manual drone monitoring is unsustainable
Existing drone surveillance required a trained operator to watch a live feed continuously. That is not scalable — human attention drifts, shifts end, and the workload is relentless across such a large area.
Security teams were always reacting, never preventing
Without automated detection, the security team only learned about threats after they had already entered or moved through the reserve. By the time rangers were mobilized, the window for interception had often already closed.
Existing tools couldn't close the gap
CCTV cameras cover fixed points. Manual drone flights depend on operator availability. Neither solution provided the continuous, intelligent, area-wide surveillance the reserve needed to protect its most vulnerable animals.
The stakes are existential
Every missed alert is a potential animal lost. With rhino populations already at critical levels, the cost of surveillance failure is measured not in inefficiency — but in irreversible loss.

Continuous monitoring, zero manual effort
The system connects to a live RTMP video stream broadcast by the reserve's drones via Google's Live Stream API. It monitors that feed continuously — 24 hours a day, 7 days a week — with no human operator required to watch the screen. The drones fly. The AI watches.
Real-time threat classification
Using YOLOv11, a state-of-the-art real-time object detection model running inside a Cloud Run container, the system analyzes every frame of the live drone feed. It classifies what it sees — distinguishing between animals, reserve staff, and potential threats such as unknown humans, vehicles, and dogs. Only genuine anomalies trigger an alert.
Instant WhatsApp and email alerts
The moment a threat is confirmed, an automated alert fires immediately. Rangers receive a WhatsApp message and email containing the detection clip, precise GPS location, and timestamp — directly to their phones. No delays, no manual handoff, no missed notifications.
Built on existing infrastructure
No new drones. No new cameras. No hardware investment. The solution integrates with what the reserve already has, using Google Cloud Storage and Cloud Run to process video fragments and route them through the detection pipeline. Deployment was fast. Disruption was minimal.
Governed by a strict ethical framework
Before a single line of code was deployed, ML6 and AMES agreed on binding ethical rules. The technology is restricted to wildlife conservation use only. It cannot be deployed on weapon-enabled or autonomous decision-making drones. It cannot be repurposed for border control or any surveillance outside conservation contexts. Responsible AI is not an afterthought here — it is a precondition.

Drones broadcast their live video feed via RTMP to Google's Live Stream API. Video fragments are captured and temporarily stored in a Cloud Storage bucket, ready for processing.

The AMES–ML6 partnership is a demonstration of what focused, ethical AI engineering can achieve when it is applied to a problem that truly matters. ML6 didn't arrive with a complex, over-engineered platform. They focused on the minimal, strategic application of AI — and delivered something immediate, powerful, and deployable in the field.

For AMES, the result is a system that works today and scales tomorrow. The same pipeline that protects Dabchick Wildlife Reserve can be replicated across any reserve with drone capability — turning a local proof of concept into a global conservation tool.

For the broader conservation community, this project proves that sophisticated AI does not need to be expensive, complicated, or ethically compromised. When technology is guided by clear purpose and strong principles, it becomes one of the most powerful tools available to protect the natural world.