DAT AI
The intelligence behind every spray decision
DAT EcoPatch is powered by deep learning that recognises individual plants in real time; telling crop from weed as the sprayer moves, and switching boom sections on and off patch by patch.
How it thinks
An accurate decision for every plant
- 1
Detect
Deep neural networks locate every plant in frame, resolving thin grass blades and tiny seedlings in dense canopy and variable light.
- 2
Classify
Each plant is separated into cash crop or weed in real time, even where weeds grow tangled through the crop.
- 3
Decide
Agronomic thresholds decide whether spraying is worthwhile and switch the right boom section on at the right moment.
Trained to tell apart
Your crop, and the weeds that matter
DAT AI separates grassweeds, broadleaf weeds, mayweed and thistles from cereal crops, with canola / oilseed rape and thistle detection currently in field trials; not just spotting “something green”, but recognising what it is. In grassland it works the other way round, picking out broad-leaf weeds like mayweed and thistle (in field trials) while protecting the grass.
Because it works at the weed-group level, you stay in control. If you already know your tank mix won't touch a particular weed, DAT can be set to ignore it, or to target only the groups you actually want to spray.
Every detection is logged with GPS, so DAT maps where each weed type sits in your field; grassweed and mayweed maps today, with thistle maps in field trials and more weed classes being added all the time.
Crops and grassland it covers
Weeds it isolates
Detection modes
Green on green, and green on brown
Green on green
Telling a green weed from a green crop, side by side and often overlapping, is the hardest problem in precision spraying. It is exactly what DAT was built to solve, in real time, and it is where our AI stands apart.
Green on brown
Picking out weeds against bare soil or stubble is more straightforward, and DAT handles it just as reliably; ideal for pre-emergence and post-harvest passes.
per plant, every decision
Fast enough to decide on the move
Optimised to run right on the machine, DAT AI detects, classifies and decides in under 12 milliseconds; real-time, green-on-green precision at working speed, with no internet connection and no lag between seeing a weed and acting on it.
Built for real fields
Accurate in the conditions you actually farm in
Fine grass and tiny weeds
Resolves thin grass blades and small seedlings that other systems miss.
Weeds within the crop
Separates weeds growing intertwined with the crop, not just isolated plants.
Any light, any soil
Robust to changing light, soil colour and reflections, so accuracy holds from dawn to dusk.
Across growth stages
Recognises grasses and weeds through their different growth stages.
The data behind the model
Trained on data that keeps growing
A robust, scalable model needs breadth and quality. Ours is built on a wide range of data gathered across different geographies and growing conditions, refined through years of field trials and testing, and it improves every single day.
Carefully labelled, every image
Models are trained on very high quality annotations. Every image passes robust checks, several times over, so each plant carries the correct class label and the training data stays clean and precise.
Freshly annotated, every day
We add newly annotated images daily, continually broadening the range of weeds, crops and conditions the model has learned from.
From field image to labelled data
0+
Unique high-resolution field images, and counting
2010–today
Collected across spring and fall seasons
Many regions
Multiple geographies and biological conditions
Every field that runs DAT contributes high-quality data, so the database grows exponentially and the model keeps getting sharper.
Want to know more?
We're happy to talk through how our AI works and what it can do for your fields.
Contact Us