Understand the scene
Label components
A supervised review grid over the components in hand (primitives, segments or classes), with a brush per class, box and lasso in the view, a bulk selector by height, size and footprint, and a baseline so the model and the human can be compared.
Labelling is free. Exporting the labels and the baseline column is Pro.
Part of Point cloud classification, Point cloud segmentation.
When to use it
After a decomposition, a segmentation or a classification, whenever the labels have to be right rather than mostly right: before a deliverable, before training data, before a count that someone will sign.
Step by step
- Open Process › Label Components…. It is available once there is something to review: a decomposition, segments or classes.
- If more than one source exists, choose which components to review under components: (primitives, segments, classes, or any integer id column).
- Choose a vocabulary: Indoor, Outdoor, Powerlines, Streetscape or Ground / off-ground. Add a class of your own with new class and + add.
- Pick a brush under paint: (keys 1 to 9 pick one). Click a tile to paint it; Ctrl-click toggles a selection, Shift-click selects a range; Enter labels the tile under the cursor; ← and → turn pages.
- In the view, Ctrl + drag draws a box and Alt + drag a lasso; right-click flies to a component.
- For many at once, open by size & place…, set height, size and footprint ranges, and select n matching; then paint n.
- Narrow the grid with refine: (to do, isolate, a kind, a class) and hide labelled. Under layers, each class can be hidden, soloed, locked so no brush can change it, renamed or recoloured.
- Press keep baseline before you start to keep the model’s labels as
classification_baseline, so model and human can be compared on export. - Press apply classification to write the labels onto the cloud now. Minimise puts the grid in the tray at the bottom left, as it is; Finish completes the step and lists the result under Results.
What it was measured at
A decomposition hands the grid 1,501 components for a 43.5 million point building, instead of the 9,626 regions region growing alone produces.
Escape, one level at a time
In the grid, Esc first drops a box or lasso in progress, then clears the selection, then clears a solo, and only then minimises the grid to the tray, so a stray key never throws away a selection you built. It is the same ladder Escape walks everywhere in the app; see Escape, one level at a time.
Limits, stated
- Closing the grid commits the labels but keeps the review state; reopening it shows what you had reviewed.
- A locked class refuses paint in both directions: nothing can be painted into it or out of it.
Related questions
How do I know how sure the classifier was?
Each classification writes a confidence column: the share of the random forest’s trees that agreed. Colour by it to see where the model hesitated, threshold on it, and review those parts in Label Components…. N3D does not certify its own output; confidence, a human override and a baseline you can export are what it offers instead.
Can I train my own classifier?
Not inside the app. You can correct labels in Label Components…, keep a baseline to compare the model against your review, and export per-segment features with n3d-render-cli classify-features (Pro) to train elsewhere.
Will the classifier certify its own result?
No, deliberately. N3D reports a per-point confidence, lets a person review and correct every component, and keeps the model’s baseline beside the human labels. That is what can be signed; a model marking its own work is not.
More in the FAQ.
Related guides
component and its primitive_kind.
Understand the scene
Classify a point cloud (four models)
Label every point with one of four trained models (Indoor, Outdoor aerial LiDAR, Powerlines, Streetscape), read how sure the model was from a per-point confidence column, and split a powerline corridor into individual trees, spans and poles.
Deliver
Export a point cloud (LAS, PLY, CSV, USD, web)
One dialog for every point cloud export - format, subsample, colours, label columns and scalar fields - with a LAS writer that puts the class in the classification byte and every id in an exact 32-bit extra byte.