Features — LiDAR Classification, 3D Objects & ML Training

Module 1

High-Performance Manual Classification

Complete 3D Selection ToolkitRectangle and lasso for fast selection, polygon for granular control, brush for irregular patterns, hull select to capture whole 3D objects from a few corner points, and smart selection that identifies objects automatically. Combine tools in multi-step selections and refine any selection by classification.
Cross-Section Profile EditorCut dynamic cross-sections through roads, terrain, and structures to inspect and edit point classes vertically.
Class Filter ProfilesInstantly toggle visibility and selectability for standard ASPRS classes (Unclassified, Ground, Low/Medium/High Vegetation, Building, Noise).
Undo/Redo Stack & Batch ActionsReclassify millions of points instantly with complete history protection.

Module 2

3D Object ExtractionTurn classified point clusters into clean, lightweight 3D objects — buildings and trees, for example.
Terrain & Contour Mesh GenerationCreate clean terrain meshes and contour lines from classified ground points.
Multi-Format ExportExport extracted objects in common CAD, GIS, and 3D formats.

Module 3

Ground Truth AnnotationUse manual tools to flag training bounding boxes and labeled sample areas.
Hyperparameter SelectionConfigure model epochs, point density subsampling, batch sizes, and feature channels (XYZ, Intensity, Return Number, RGB).
Local CPU AccelerationHarness your host's multi-core CPU for local model training — no GPU needed, and zero data leaves your network.
Automated Batch ProcessingApply trained classification weights across entire folders of LAS/LAZ tiles automatically.

Feature	Specification
Supported File Formats	LAS, LAZ, COPC (Cloud Optimized Point Clouds), EPT, PLY
Rendering Engine	WebGL accelerated point rendering engine
Point Capacity	Up to 100M+ points handled via dynamic level-of-detail (LOD) octrees
Data Privacy	100% On-Premise / Local network processing