Can AI camera traps catch wild dogs before they reach the sheep?

Testing whether an AI camera-trap alert is worth building for graziers and doggers, before writing any code.

Wild dogs (dingoes and hybrids) get past or through fences and attack sheep, often at night across large paddocks. Many kills are found late, at lamb marking or shearing. A federal estimate puts losses at $302M a year.

Doggers (professional trappers) need to know where dogs come in and when, so they can place traps and baits. Today that comes from tracks, sightings, carcasses and trail cameras that must be checked by hand. Most camera triggers are kangaroos, foxes, cattle or wind.

A camera trap that recognises wild dogs and sends an alert within hours, so a dogger can act on where dogs are actually coming through.

• 6 rounds of desk research: losses, existing tools, regulation and connectivity in remote areas.
• Prior attempts mapped, including a 2018–19 research prototype with no commercial product found since.
• Existing open models and Australian camera-trap datasets shortlisted, with licences checked.
• Interview guide written; first outreach sent to doggers, wild dog coordinators and councils.

• Does a faster alert change the outcome, or is the response days away anyway?
• Who acts on the alert: the grazier, the dogger, or a local committee?
• Who pays, and how do alerts get out where there is no mobile coverage?

Interviews with graziers and doggers, and a quick test of an existing dog-detection model on public Australian camera-trap images.