Classes: Automatic Object Recognition on Your Orthophoto

A fresh orthophoto is a fine picture, but it is not an inventory: to count the vehicles on site, add up roof areas or find out how much ground the roads cover, you would have to trace every object by hand. Classes does that for you — on an orthophoto produced in the CartaX cloud, buildings, roads, vehicles, water and more are recognised automatically; you click on the map and read the count and the area.

What is recognised?

Class	What it captures
Building	Roof footprints and structures
Road	Asphalt and gravel road surfaces
Vegetation	Trees, shrubs and grass
Water	Ponds, channels, pools
Vehicle	Cars, trucks, construction machinery
Human-made	Other artificial objects: containers, concrete pads, platforms
Farmland	Cultivated and ploughed fields
Power line	Power and communication lines
Custom	Extra objects you ask for (below)

The Custom class is built from your own description. When you start cloud production, the Photogrammetry card in the Upload data dialog has an Extra objects (optional) field: type the objects you want found — greenhouse, silo, wire fence — in English. Whatever is found arrives on the orthophoto as the "Custom" class.

Opening the Classes window

In the row of an orthophoto layer that has class output, a Classes button appears right next to the three-dot menu. The button shows only on orthophotos CartaX produced with classes — in the cloud, or on this computer with CartaX Studio 0.6+; an orthophoto you uploaded yourself does not have it.

The button opens a small, draggable window; grab its top edge and move it anywhere over the map. For every class you get:

• a colour swatch — click it to recolour that class; your choice is remembered in this browser and only affects your own view (it never touches the data or a share),
• a show / hide switch,
• the count (how many objects were found),
• the measure — for area classes (building, vehicle, farmland…) the total area (m² under one hectare, ha above); for linear classes (road, power line) the total length (m under a kilometre, km above).

The area or perimeter of a road tells you nothing useful — what you want to know is how many metres of road there are, so both the window and the badge switch to length for linear classes.

Source: point cloud or orthophoto

Some productions show a small Source selector at the top of the window: Point cloud and Orthophoto. Both come from the same AI labels; what differs is where those labels are projected.

• Point cloud — classes are derived from the 3D points. Outlines sit on the terrain surface and stay faithful on bulky objects.
• Orthophoto — the same labels are projected straight onto the surface model. This catches noticeably more thin objects (power lines above all, and small vehicles), because a cable is hard to reconstruct in 3D yet plainly visible in the photograph.

Orthophoto is the default. Measured on the same site: power lines go from 217 m to 918 m, and vehicle and container outlines land more accurately. The difference is structural — a cable is hard to reconstruct in 3D yet plainly visible in the photograph. The point cloud source stays available: where the surface model has gaps, the orthophoto source has no labels there, and the second source earns its keep.

The selector appears only when both outputs exist. Switching hides the layers currently drawn and redraws them from the new source, so you compare the two side by side rather than stacked on top of each other.

The total area at the bottom sums the area classes only (roads and power lines are excluded — adding a length to an area produces a meaningless number). Classes start hidden, so the map does not fill with coloured polygons until you ask.

Exploring by clicking on the map

The quickest way in is to click. With the window open, click an object on the map — a container, a vehicle, a building:

1. If that object's class is hidden, it becomes visible automatically.
2. A badge appears on the object.

What the badge shows depends on the kind of object:

• Area classes (building, vehicle, human-made, farmland, water, vegetation): "Building · 1,240 m² · Perimeter 148 m". If the shape is simple — a four-, five- or six-cornered footprint — each edge length is listed too.
• Linear classes (road, power line): "Road · 254 m (≈2.8 m wide)". No perimeter, no edge list.

Automatically traced polygons usually carry hundreds of vertices (we measured about 190 on an average road), so printing that list would be noise rather than information. Edge lengths therefore appear only on genuinely simple shapes. Close the badge with its close button.

It stays out of the layer list and out of your measurements

Classes are not added to the layer list as vector layers. That is deliberate: measurements are attached to layers, so hundreds of automatic polygons would bloat the list and get mixed into your shares and measurement records. Close the Classes window and the map returns to its previous state; your measurements remain untouched.

Classes in the point cloud

The point cloud from the same production is classified too. The Classification panel of the point cloud viewer lists Ground, Vegetation (low / high), Building, Road surface, Water, Vehicle, Human-made and Wire conductor in their own colours, each with its own Show / Hide. Hiding the vegetation to leave only the ground, or showing buildings alone, is a single click.

Where it helps

• Site inventory — how many vehicles are on site, how many structures are standing: open the window, read the counts.
• Area breakdown — report vegetation, road and building areas in hectares.
• Presentations — show the client the "buildings blue, roads grey" view live, and click to reveal an area.

How do you get it?

Classes come with cloud photogrammetry production on the Professional plan; there is no separate button and no extra charge. Local production with CartaX Studio and orthophotos uploaded from elsewhere do not get Classes.

Related guides

• Cloud photogrammetry production
• Local production with CartaX Studio
• Viewing point clouds in the browser
• Plans and tokens

Frequently asked questions

Question	Answer
I cannot see the Classes button.	It appears only on orthophotos CartaX produced with classes. You must have ticked Segmentation for that run; in Studio that also requires an NVIDIA graphics card. Orthophotos you uploaded yourself do not have it.
Why are classes hidden at first?	To keep the map clean; clicking an object reveals its class automatically.
Can I edit the classes?	The results are automatic; you show and hide them from the window.
Does it affect my measurements?	No; classes never enter the layer list or your measurement records.
What should I type as a custom object?	The name of the object you are looking for (greenhouse, silo, wire fence), in English.
Is it in the point cloud as well?	Yes; every class can be toggled separately from the Classification panel.
Does it cost extra tokens?	No; it is included in the price of cloud production.