Aerial crop scouting, built for operational clarity
agri-edge is a software prototype that turns drone imagery into field-level scouting data and treatment recommendations. Simulated and bench-tested.
Farming this land yourself?
Where this stands
Read the roadmapSoftware prototype
Simulated and bench-tested.
88/88 simulation requirements
§§79–166. Not field-complete.
Not yet
Nothing here is a field result.
Read-only toward the aircraft
The Pi never flies.
The problem
Aerial scouting tools stop at pictures, or act without enough human oversight. agri-edge is built around the step in between.
Scouting is slow and manual
Walking or driving a field to find problems takes time that operators do not always have.
Imagery is not a decision
Drone pictures pile up. Turning them into reviewable, field-level findings is the missing step.
Autonomy without oversight is hard to trust
Systems that act on their own are hard to approve and hard to insure. Review and limits have to be designed in.
Scout
Turn aerial imagery into field-level observations.
Plan
Define the field, flight lines and geofence before anything is recommended.
Review
Check detections and alerts before a recommendation is made.
The evidence so far
Published numbers only. None of them is a field result.
Full-loop and field-hardware validation have not been completed.
Safety and risk discipline
The limits are designed in, not added later.
The Pi never flies
The companion computer has no flight authority. It observes and analyses; the pilot and flight controller fly the aircraft.
MAVLink is read-only toward the aircraft
The link carries telemetry in. Nothing is sent to the aircraft by agri-edge software.
Treatment needs validation and a geofence
A treatment recommendation requires validated_locally and a defined geofence before it can be issued.
Phone companion stays read-only
The phone companion at /app shows status and results. It cannot command anything.
The honest stage
A software prototype, simulated and bench-tested. The next stage is validation, and it has not started.
What we do not claim
- No field results are claimed.
- Hailo/Pi, the Detection API and the SITL adapter are not yet marked validated.
- weed_crop 0.749 is a pipeline proof only, not field-aerial.
- No customers, certifications or approvals are implied.
Talk to us about a pilot
Contact us for pilot pricing. We will scope an evaluation or pilot with you.
What it looks for
Written for the person who walks the field. A drone flies over, the software looks at the pictures, and it points you to places worth a closer look.
Weeds
Where weeds are showing up between your rows, so you do not have to find them on foot.
Patches that look different
Areas of the crop that look different from the rest of the field, so you know where to go and check.
Counts
A count of what is in view, such as animals in a paddock, to save you a head count.
Today this is a software prototype tested on public image sets and in simulation. It has not been tested in real fields yet.
What you see
- A map of your field with numbered spots.
- A picture for each spot, so you can see what the software saw.
- A short list, not a wall of data.
What you decide
- Whether a spot is worth a visit.
- Whether anything needs doing, and when.
- Or to ignore it. It is your field and your call.
You stay in control
The software never flies the drone
It reads what the aircraft reports and never sends it commands. The pilot flies.
Nothing is suggested outside your boundary
A treatment suggestion only appears once the area has been checked and sits inside a boundary you set.
Your phone only shows results
The phone view is read-only. It cannot start, stop or change anything.
A person reviews every suggestion
The software recommends. You approve. Nothing happens on its own.
Curious how it looks? The demo uses synthetic data and a made-up field. Want to talk about trying it on yours? We will scope a pilot together.
Talk to us about a pilot
Contact us for pilot pricing. No obligation to discuss.