🚨 New publication 🚨 Michael Walsh & Stephen Powles PhD FAA, FTSE look at the role of harvest weed seed control in Australian grain production systems. What does it mean for weed recognition? Read the full paper here or summary below from WeedSmart Sydney Agriculture publish.csiro.au/CP/pdf/CP21647

Images are everything in #deeplearning! E.g. #ImageNet has >14 M images in 20k classes. Nothing of the same scale exists for weeds (Weed-AI is trying!) but there are open datasets. For weed recognition, here's a 🧵of all weeds datasets I've found. 📸🌱🧵

🚨🌱new dataset! If you're interested in a project for broadleaf weed recognition in your lawn, this might be a good place to start. 567 instances of weeds (rather unbalanced) 78 images Perfect to get a YOLOv5 model up and running.

weed-ai.sydney.edu.au is an open data platform that the Sydney Informatics Hub is building to enable R&D into computer vision methods to identify and control weeds. It's great to see new data coming into the system!

A university of OWLs, showing the changes in design since way back in 2019 when I first built an enclosure. If you want to learn how to build an OWL & see a live demo, it's all happening on Wed May 4th, 12 - 1:30pm AEST. Register here: eventbrite.com.au/e/openweedloca… William Salter

We’re about to start our OpenWeedLocator live assembly and demonstration with Guy Coleman at the helm 🎥🎙

Had a great time giving this OWL assembly/demo with William Salter & Gus! Hopefully demonstrated the whole development process: parts to paddock. Glad nothing too major went awry!

Spot the difference: One collected at 5km/h, the other at 30km/h, dust and all. Both with the #OpenWeedLocator software.

🚨🌱New dataset! This 552 image dataset of wild radish in wheat is the first on Weed-AI that makes use of the new weed growth stage/plant part specification. Plus Weed-AI now has dataset editing & version control functionality too. Sydney Agriculture USyd Core Research Facilities

The first dataset with fireweed was just uploaded to Weed-AI! A small dataset of fireweed at both vegetative and flowering growth stages collected with the Sony Alpha A7R IV A, providing 61MP per image.

Weed-AI now hosts: 🗄 17 datasets, 20,891 images 🌾 8 crops + fallow 🌱 28 species + poaceae/broadleaf groups 🌏 datasets from: 🇦🇺🇩🇰🇦🇷🇺🇲 If you've got datasets of weeds or need data for a project, check out Weed-AI weed-ai.sydney.edu.au

Guy Coleman Great to see them go out to a good home👍🏼 one rig working and another 3 coming

We're looking for growers to join a beta testing group Weed-AI, for feedback on the user experience - you'd just need to have images to annotate/upload. We'll form an ongoing group around opensource weed recognition too. Let me know if you're interested! Dr Henry Lydecker

Guy Coleman & Weed-AI are seeking growers to join a beta testing group for feedback user experience. 📷 All you need are the images to upload. DM Guy Coleman if you're interested in participating x.com/GeezaColeman/s…

Really enjoyed writing this article for Tom Wolf and Jason Deveau! Check it out if you're interested in the tech behind camera sprayers, where it all started and how an #opensource approach will help it succeed. Sydney Science Sydney Agriculture WeedSmart

This machine’s algorithm is capable of distinguishing between #BroadleafWeeds, #GrassWeeds, and #ginger with a great level of accuracy and is helping to detect and spray for #weeds in a ginger crop. To read more about this project, visit bit.ly/3o0cmJK

To pass the 100th commit🥳for the #OpenWeedLocator, we've released an improved image sampling tool. Now you can save whole images, cropped bounding boxes around the detections or cropped squares. To collect images - just plug in a USB, set it up and go! github.com/geezacoleman/O…

Just finished assembling the compact OWL or OWLet, after some great feedback during the Mingenew Midwest Expo. Fits inside 125 x 75 x 3mm RHS steel for smash protection, same features as the big OWL. Designs/assembly guide to follow!

The OpenWeedLocator demo unit will be at the Mallee Sustainable Farming Future of Weed Detection Expo in Karoonda next week. A great opportunity to get hands on experience with how image-based weed detection works and DIY units. Kylie Matthewson AgriFutures Australia

About 50 years ago, the tech behind reflectance-based 'green-on-brown' sprayers was published. Since then a lot (!) has changed. We went through ~175 papers from the last 50 years of weed recognition research for this review. cambridge.org/core/journals/…