Guy Coleman

An OWL set up for weed detection and spot spraying trials on an Amiga robot by Texas A&M PhD students Bholu and Vijai.

First of its kind on this robotic platform

#opensource #agsky

Please pass along, I’m recruiting PhD students to join our Macroecology Lab @uofa-eeb.bsky.social We study phys ecology, macroecology, biodiversity - spanning scaling, trait-based ecology, theory, comparative biology & ecoinformatics. Several avenues for funding. Please reach out if interested🧪🌐🌾

Congratulations!

Thanks for listening and sharing the link!

Some really interesting talks.

First up is AI-based weed detection for precision agriculture: OpenWeedLocator presented by @geezacoleman.bsky.social
Check out all the details 👇
www.nature.com/articles/s41...

Context for the experiment here:

bsky.app/profile/geez...

Photoshoot time.

Highest mimetic vs. least mimetic individuals pretty distinct.

#agsky #science

Today, I have been working on a controller for the Giant Unicorn Hat. Have written a function to move the powered fadder to a set position, using a @pimoroni.com motor 2040, running CircuitPython to monitor/control the faders and the NeoTreilis.
#GiganticUnicorn

Thank you!

Open to ideas on what to do with about 10,000 images of 3500 ryegrass plants besides this experiment, seems like it could be useful?

Next steps are to use RootPainter to correctively annotate and train a unet model go extract the plant from the background, then finally run all extractions through the classifier and see how 3 years of work went.

Sweaty palms when making the first ggplot.

Every plant imaged three times over three days to get a mean model prediction. Using a Sony Alpha A7IV for imaging.

Would absolutely automate this for a future experiment...

#agsky

Differences between the third generation of high and low wheat similarity ryegrass are pretty clear through the camera lens.

So much more reflective leaf surface facing the camera, compared to the wheat-mimetic.

All 3 selections made by a wheat-ryegrass classification algorithm. No human input.

Looks very pretty with the randomised design and colourful labels now.

First time seeing the F3 high similarity population - exciting times

Interim results from the PhD would suggest that, yes, image classification algorithms can drive morphological changes in plants towards a mimetic phenotype.

The final 720 pots in a three year experiment.

Can weeds evolve to avoid being detected by cameras?

Pretty cool to see targeted steam/hot water applications using presumedly WeedITs or similar here in 🇩🇰.

Have oversaturated the photo so the blue line is more obvious. Very compact machine.

#agsky

Postdoc fellowship in One Crop Health simulation, mechanistic co-simulation of Insects, Weeds and Plant Infestation in Agricultural Fields

Should also mention there is a three year postdoc position available as well as part of the same project

#sciencesky

employment.ku.dk/faculty/?sho...

The OneCropHealth project is looking at the interactions of all crop management processes and trying to implement better precision agriculture tools.

A good mix of hardware, software and field work is likely.

Get in touch if you're interested!

#agsky

2 PhD fellowship in One Crop Health Computer Vision for Monitoring of Pests, Weeds and Beneficials in Agricultural Fields and in simulation, mechanistic co-simulation of Insects, Weeds and Plant Infes...

Come work with us at the University of Copenhagen on image recognition problems in agriculture!

Some really interesting projects that go across institutions and borders.

employment.ku.dk/phd/?show=16...

New era for WA cotton industry as $60m processing facility opens The opening of the Kimberley Cotton Gin means growers will no longer need to transport their cotton more than 3,500 kilometres to Queensland for processing.

The opening of the Kimberley Cotton Gin means growers will no longer need to transport their cotton more than 3,500 kilometres to Queensland for processing.

Hard at work hardfacing our points and leading edges! 💪 Our robots ensure top-quality results, so you get the best performance every time.

#AgSky

It's in working condition currently over on the wireless-display branch if you want to try it out:

github.com/geezacoleman...

A few bugs to sort out and documentation to add before merging

Finally got the MQTT-based operation of the OWL to some form of usefulness.

Only takes two lines to set the whole thing up:

git clone
bash owl_setup.sh

The interface lets you monitor, control and download data to check its quality.

Just launched the new WeedAI web app!

Check it out here: weedai.app

Visualise the 400k+ annotations on over 41k images from around the world.

Anyone can add data too - upload your dataset and fill in the gaps on the map.

Victorian Country Hour - ABC listen Red meat abandons it's plan to be carbon neutral by 2030, South Australia delays the fruit sticker phase out and we crack some nuts

The red meat industry is walking away from its goal of being carbon neutral by 2030.

It’ll be replaced by a methodology that will take a year to develop.

The Red Meat Advisory Council, government and Farmers for Climate Action give their thoughts.

Listen: 👇🎧

www.abc.net.au/listen/progr...

It's weirdly nostalgic finding a @raspberrypi.com 3B+ from 2018 with my very early days Python code on it.

self.self = self

I guess? This Pi will never be updated.

An OWL building workshop with 11 farmers from Western Aus!

Lots of great feedback on the build process.

Close up of the board itself. Wiring was definitely a hassle, but decided for extra connectors so the lasers wouldn't always be stuck in this.

Set up a 16 channel laser demo for the OpenWeedLocator using the new 16 channel control board.

Plan is to make precision weed control in horticulture more accessible.

Box next to the OWL controller.

I think learning how to make PCBs and not spending 2 days wiring this would be more productive...