Cam IncollFieldwalker uses AI to automatically identify trees, buildings, water features, and vegetation on aerial images — turning what used to be hours of manual tracing into a starting point you can work with in seconds.
When you start a new site design, the first real task after getting your base map is figuring out what's already there. Every tree. Every building. The dam in the back corner. The dense patch of vegetation along the creek line. The fence lines, the access tracks, the infrastructure.
On a suburban block, that might take twenty minutes of careful tracing. On a ten-acre property, it's hours. On a farm, it's the kind of task that makes you put the whole project aside for the weekend and never come back to it.
And all of it is mechanical work. You're not designing. You're not observing in the permaculture sense — reading the landscape, understanding relationships, noticing patterns. You're tracing outlines over a satellite image. It's necessary, but it's not the work that matters.
What if the tracing was already done?
This is where AI detection comes in. Fieldwalker can look at your aerial image and identify what's on it — trees, buildings, dense vegetation, water features. It outlines them, classifies them, and presents them for your review.
The way it works in practice: you tap a feature on the map, and the AI identifies what it is. It draws the outline, tells you what it thinks it's looking at — a tree, a building, a patch of dense vegetation — and gives you the option to confirm, correct, or dismiss.
It's fast. What takes hours by hand takes seconds per feature. And because the AI is identifying and classifying in one step, you're not just getting outlines — you're getting categorised features that the tool understands.
From detection to design
This is the part that matters most. Detected features aren't just annotations sitting on top of your image. They become design elements.
A detected tree becomes a tree symbol — rendered in the illustrated permaculture style, sized to match the canopy visible in the aerial image. A cluster of trees that the AI identifies as dense vegetation becomes a canopy blob — that single amorphous shape you'd draw in a Mollison-style design, with the scalloped organic edge and trunk dots within.
A building becomes a building symbol. A body of water becomes a water feature. The detection gives you the raw identification; the conversion step translates it into the visual language of permaculture design.
This means your base map goes from "aerial image with some outlines" to "the beginning of a design document" in minutes. The existing trees are already placed. The buildings are already drawn. The dense vegetation is already rendered as canopy. You're not starting from a blank canvas — you're starting from a site that already has its existing elements mapped.
It's a starting point, not the final word
I want to be honest about this: AI detection isn't perfect. It's very good at trees and buildings. It's solid on dense vegetation and water. It can miss things, especially in complex landscapes where features overlap or where the image quality is uneven.
That's fine. The detection is a starting point — it saves the bulk of the tracing work and gives you a head start. The designer still reviews, corrects, and refines. You might need to add a tree the AI missed, or reclassify something it got wrong, or adjust an outline that isn't quite right.
This is actually how it should work. The observe-and-refine step is part of the permaculture design process. You're looking at your site, checking what's there, confirming your understanding. The AI just means you're refining a draft instead of starting from nothing.
And every correction you make feeds back into the system. Over time, the detection gets better — learning from real corrections on real sites. Your input improves it for the next designer.
Spending time where it counts
The reason this matters isn't the technology. It's what it frees you up to do.
If you spend your first two hours tracing trees, you're tired before you start designing. If the trees are already there — roughly right, ready for review — you spend those two hours thinking about where the swale goes, how the zones relate to the existing vegetation, whether that dense patch along the creek should be preserved as a windbreak or thinned to open up a view corridor.
That's the work. That's what permaculture design actually is — reading the landscape, understanding the relationships, making decisions about how to work with what's already there. The mechanical parts should take as little time as possible so you can spend your energy on the parts that require your judgment, your observation, and your design thinking.
Fieldwalker's detection is built around that principle. Handle the mechanical work fast. Get the designer to the interesting decisions faster. Because the interesting decisions are the ones that make the design good.