Auto-Detect scans your entire map image and identifies trees, buildings, water features, forest and cleared areas — drawing everything it finds straight onto your map, ready for you to tune with five simple controls.
Auto-Detect looks at the whole map in one pass, places what it finds, and gives you category-level controls to dial the results in at your own pace.
The new Auto-Detect is live in V1.0 — based on feedback from users, we have completely rebuilt the auto-detect experience. Point Detect has been retired: the new Auto-Detect covers its ground with less effort.
Quick start
Your satellite image contains lots of useful information about the location of trees, buildings, water and vegetation. Fieldwalker's AI can identify these automatically, saving you hours of manual tracing.
- Open the Design tab and click Detect features in the Auto-Detect panel.
- The AI scans the entire map and draws everything it finds straight onto your map.
- Five category rows — Trees, Forest, Buildings, Water, Cleared — each have a Less–More slider and a show/hide toggle to tune what stays.
- For object-level control, open Fine-tune objects →.
- Press Add detected objects to map to turn everything into real, editable design objects.
Starting a scan
Open a map in the editor and go to the Design tab. The Auto-Detect panel is at the top.
Click Detect features to start the scan. It runs once per map — there's no re-run option, because the results are cached and kept.
You can keep working while it runs — editing, panning, using other tools. The panel shows progress in the background.
How long it takes
A typical scan completes in well under a minute. Allow up to a few minutes for larger maps or if the detection service is waking up from idle.
You don't need to stay on the page. Navigate away, close the tab, even reload — when you come back, Auto-Detect picks up the results automatically or resumes showing progress if it's still running.
What it finds
Auto-Detect identifies five categories of features:
- Trees — individual tree crowns, counted one by one
- Buildings — structures on the site
- Water — dams, ponds, water bodies
- Forest — continuous dense canopy, treated as an area and measured in hectares
- Cleared — open cleared ground, also measured in hectares
Roads and paths are excluded from Auto-Detect results — they were unreliable in testing. Use the design tools to add paths manually.
Tuning the results
When the scan finishes, everything found is already on your map, and the panel's summary line tells you what's there in your units — for example, "1,150 objects · 15.5 ha marked on your map". Counted features (trees, buildings, water) are reported as counts; area features (forest, cleared) as hectares.
While you're tuning, the map image dims slightly so detected features stand out. You can switch this off with the Dim background toggle.
The category rows
Each of the five categories gets a row with:
- A Less–More slider — drag towards Less to keep only the most confident detections on the map; towards More to show everything found. The map updates live as you drag.
- A status line that tells you exactly where you stand — "Showing 1,089 of 1,142 trees" while there's more available, or "Showing all 1,142 trees found." when the slider has nothing more to give. Area categories read "Marking 12.4 ha of forest".
- A show/hide eye — hide a whole category from the map (and from the summary figures) without losing it.
Fine-tune objects
For object-by-object control, open Fine-tune objects → at the bottom of the panel.
Objects are grouped into "Detected · on the map" and "Pending · not on the map". Each row shows the category and confidence. Keep moves a pending object onto the map; Remove takes a detected object off it (it stays listed as pending — nothing is deleted).
The list and the map are linked both ways: hover a row and the object glows on the map; hover an object on the map and its row highlights in the list. The crosshair button on a row centres the map on that object.
Selecting multiple rows: click to select, shift-click to extend. A bulk bar appears with Keep / Remove for the whole selection at once.
Correcting a detection
Hover any detected object on the map to see what the AI thinks it is and how confident it is — for example "Tree · 47%". Click it to open a small card where you can:
- Correct the category if it guessed wrong
- Tick "Don't add to map" to drop it entirely
- Optionally note what went wrong — this feedback helps improve future detection
Adding to your design
Add detected objects to map is the commit step. Everything currently showing on the map becomes real, editable design objects — the same kind you'd get from placing a symbol by hand. Move, restyle or delete them like anything else.
If any category is hidden when you commit, you'll be asked: "Add hidden categories too?" — choose to commit just the visible ones, include the hidden ones as well, or take a look at what's hidden first.
After committing, the panel collapses into a summary — "1,150 objects · 15.5 ha added" — with a note of anything still pending, e.g. "14 pending objects kept for later review".
Changed your mind?
- Review pending on the summary card reopens the fine-tune list showing what wasn't added. Keep more and commit again — only the newly kept items are added; nothing already committed gets duplicated.
- Category rows also show a Review → shortcut whenever there's something pending behind them.
- To remove something you've already added, delete it the same way you'd delete any other design object.
What to expect
Detection quality depends on the imagery. Sharper, more recent aerial images give more reliable results than older or lower-resolution captures.
A few things to be aware of:
- Dense, tangled bush is detected as a Forest area rather than individual trees — that's by design, not a limitation.
- Scattered farm-paddock trees, especially widely spaced individual trees, can sometimes be missed by the full-map scan. Add any stragglers with the design tools.
- Round structures like water tanks can occasionally be classified as either a building or a water feature. Worth a quick visual check on round objects.
- If nothing clears the confidence bar, the panel says "Nothing found confidently."
- If imagery is flagged as too low-quality for reliable detection, Auto-Detect may be unavailable for that image.