YOU CAN BUILD THE CAMERA THAT SPRAYS ONLY THE WEEDS
A Raspberry Pi watches the ground pass under the boom and opens only the nozzle above a weed.
by Guy Coleman, William Salter, Angus Macintyre
GardenAutomation
Built withRaspberry Pi3D printing
- difficulty
- ●●●●○
- time
- a weekend-plus
- license
- MIT
- repo
- repo ACTIVE501 stars
●●●●○ · a weekend-plus · MIT · 501 stars · repo ACTIVE
WHAT YOU’LL NEED
Jump to section
COMPAREE VERDICT
OpenWeedLocator is a camera unit that mounts on a boom sprayer and fires only the nozzle above a weed, instead of blanketing the whole field. The hardware is a Raspberry Pi (a Pi 5 in the current parts list) with a Raspberry Pi camera (Global Shutter recommended), a relay board and a 3D-printed or aluminium enclosure. Detection runs on the Pi: green detection for weeds on fallow ground out of the box, and an optional Green-on-Green mode with a trained model for weeds inside a crop, which adds about 2 GB of software. The project is MIT licensed, documented on a dedicated docs site, backed by peer-reviewed papers from the University of Sydney team, and actively maintained. The most likely failure point is calibration: lighting and soil colour vary, so plan time in the field tuning the detection thresholds. If you have a boom, the mechanical skills to mount the unit and the patience to dial it in, this is a weekend-plus build; if you expect plug-and-play computer vision that works everywhere, wait.
IN THE REPO
GOOD TO KNOW
- —Enclosure STLs, relay board wiring diagram, full parts list and installation script are all present.
- —The colour-based fallow detector is ready; the green-on-green crop model is still being refined.
- —A guided assembly and field demonstration video from the authors' lab is on YouTube, so you can watch it work before building.
- —MIT licensed and permits commercial use.
- —You supply your own boom, tractor and hydraulic solenoid valves—this is the computer vision add-on.
- —The authors are actively maintaining the project; last push was this month.
Parts to buy
6 itemsFrom our check of the build. Exact quantities and part numbers are in the creator’s BOM.
Can I build this?
Build at your own risk. Projects involve tools, electronics and sometimes mains voltage — follow the creator’s safety notes.
Videos
OpenWeedLocator: Guided Assembly and Demonstration1:13:34
Full assembly from parts to enclosure, then live field test behind a ute; filmed by the authors
More builds like this
All projectsGallery
Start here
Navigation into the creator’s own docs — we don’t rewrite the guide, we route you to the source.
- 1.Watch the assembly video (See the whole build and field test before ordering parts)
- 2.Read the BOM and check sourcing(The docs' parts list covers the Pi, camera options (Global Shutter recommended), relay HAT, connectors and solenoids.)
- 3.Print the enclosure(STLs for the camera housing and mounting brackets)
- 4.Install the software(Scripted Pi setup; the guide walks through every dependency)
Resources
Documentation, files and community threads for this build — we link straight to the original sources and never rehost the creator’s files.
KNOWN ISSUES
- The system works in the video because the authors tuned the HSV thresholds for that soil and that light; your field will need its own calibration session.
- The green-on-green crop detector is the hard problem and is still being refined—fallow detection is the production-ready mode.
- You need solenoid valves that match your boom's hydraulics; the relay board controls them, but sourcing and plumbing them is on you.
- Mounting the unit so the camera sees the ground clearly while the boom bounces over ruts is harder than the electronics.
- Detection performance depends on speed and conditions; the authors published speed tests up to 30 km/h, so test at the speed you will actually spray.
- If your tractor cab does not have line-of-sight to the Pi, you will need a longer cable or a wireless link to monitor detection live.
Does it work for weeds inside a crop, or only on bare ground?
The fallow mode separates any green from soil and is ready to use. The green-on-green crop mode uses a trained model and is still being improved; it works, but accuracy varies by crop and weed type.
Can I run this on a Pi 3?
The parts list names a Raspberry Pi 5 (4 GB) as the main choice, with a Pi 4 or 3B+ as supported alternatives. Test at the speed you will actually spray.
Do I need one unit per nozzle, or one for the whole boom?
One unit drives several nozzles through its relay board; the standard build uses a 4-channel relay HAT, and the README shows larger builds such as a 16-channel vegetable OWL.
What happens if the camera gets mud on the lens?
Detection stops working. The enclosure keeps rain out, but you will need to wipe the lens between passes if conditions are muddy.
Community builds
No community builds yet — be the first, we feature the best ones.
Discussion1
FROM THE COMPAREE TEAM
Out of the box, OWL spots green weeds on bare fallow ground; finding weeds inside a growing crop needs the optional Green-on-Green setup with a YOLO model and about 2 GB of extra software. If you were building this, would you start with fallow, or go straight to training a model for your crop?
Guy Coleman, William Salter, Angus Macintyre
Built by weed control researchers at the University of Sydney. The project started in 2021 and is backed by peer-reviewed papers, including tests of fallow weed detection at speeds up to 30 km/h.
Keep building — related projects
Same category or platform, picked by how close they are to this build.
NO CAD FILESYour Raspberry Pi can run a hydroponic or mushroom farm on its own, holding the conditions while you harvest
GardenFor intermediate builders
FULL CADTurn your old TV into 90s cable
DisplaysFor intermediate builders
FULL CADYou can build a robot that hangs from your ceiling and picks things up off the floor
RoboticsFor expert builders
FULL CADYou can build a real Harry Potter wand that controls your home
AutomationFor advanced builders
NO CAD FILESFree software and a USB adapter turn a laptop into a full DMX lighting desk
AudioFor intermediate builders
NO CAD FILESYou can make the Furby Hasbro abandoned say anything you want
AutomationFor intermediate builders
DISCLAIMER
- Comparee is not the author of the projects featured here. All rights to each project belong to its creator — every page links to the original source, and we never host creators’ files.
- Information is provided without warranty and may become outdated as projects evolve. Prices are indicative bands only — always check the creator’s parts list for current costs.
- Building and operating any project is at your own responsibility. Protective equipment, safe workshop practice and compliance with local regulations are the builder’s responsibility.
CompareeTEAM1mo agoedited
Practical notes from our verification: the repository has been actively maintained since 2021 and was last updated in September 2026, the guided assembly video is hosted on the University of Sydney Precision Weed Control Lab YouTube channel, and the documentation site covers hardware assembly, wiring, software install and controllers, with a community forum alongside. The single biggest time sink will not be the build — it will be the time you spend in the field tuning detection for your soil colour and lighting. The authors published peer-reviewed papers alongside the code, so the detection pipeline is documented at both the maker and the academic level. One thing to plan for: the relay board switches 12 V devices up to 10 A per channel, and while the docs recommend solenoid valves for spot spraying, you still need to check that the valve ratings match your boom's operating pressure. Correction (4 October 2026): we re-checked this page line by line against the project's own repository, documentation and videos, and fixed errors in earlier versions.