# Neurones 3D Science: 20 live 3D demos > 20 real 3D projects, each replayed live in the browser by Neurones 3D, the 3D data engine of Florent Poux (3D Geodata Academy): point clouds, LiDAR, laser scans and photos turned into BIM models, floor plans, tree inventories, solar studies and trained 3D AI. Nothing to install: each demo is the app itself, replaying what it computed on real example data. Gallery: https://learngeodata.eu/n3d-demo/ The five pillars: Smart Buildings, Safety and Compliance, Forestry and Environment, Energy and Utilities, 3D World Models (AI). ## Smart Buildings Scans become BIM models, floor plans, asset registers and quotes. - [Point cloud to BIM (IFC)](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/scan_to_bim/index.html): From a laser scan of an office to an IFC4 model: walls, doors, windows, spaces, furniture. Measured: 5 walls, 3 doors, 3 windows; 3 spaces, 41 pieces of furniture. - [Scan to floor plan](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/scan_to_floor_plan/index.html): From a scan of one room to a dimensioned floor plan, as PDF, SVG or DXF. Measured: 4 walls, 31.0 m; 1 room, 55.2 m²; 1 door, 30 pieces of furniture. - [Asset inventory from a 3D scan](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/asset_inventory/index.html): From a scan of an office to every object numbered and measured, room by room. Measured: 57 objects; 3 rooms; 4,166 EUR of furniture. - [Quantity takeoff and renovation quote](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/renovation_quote/index.html): From a classified office scan to quantities per room and a priced quote. Measured: 3 rooms; 7,017.20 EUR incl. VAT. - [Kitchen fittings from LiDAR and photo fusion](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/multimodal_fusion/index.html): From a phone scan of a kitchen with its photos to every fitting named and modelled in IFC. - [City digital twin from LiDAR HD and open data](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/territory/index.html): From IGN LiDAR HD and French public registers to buildings, trees, roads, parcels and their facts. Measured: 406 buildings, 1,382 trees; 422 parcels, a sheet each; 295 flags: scan vs registers. ## Safety and Compliance Accessibility audits, escape routes, and what changed between two visits. - [Accessibility audit from a 3D scan](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/accessibility_audit/index.html): From a scan of a floor on three levels to every barrier for a wheelchair, numbered. Measured: 36 issues for a wheelchair; 5 of 22 rooms reachable. - [Evacuation route planning](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/evacuation_plan/index.html): From a scan of an office floor to the exits and the way out of every room. Measured: 36 rooms; 10 exits; worst route 23.5 m. - [Change detection between two scans](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/change_detection/index.html): From two scans of one office to every object moved, turned, removed or added. Measured: 6 objects changed: 3 moved, 1 turned, 1 removed, 1 added; aligned to 0.5 cm. ## Forestry and Environment Every tree found, measured and modelled, from a drone, a plane or on foot. - [Tree detection from drone photogrammetry](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/trees_from_drone/index.html): From a drone photo survey of a forest to every tree: its crown, its height, its model. Measured: 73 trees; 378 per hectare; tallest 19.2 m. - [Individual tree segmentation from LiDAR](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/trees_from_lidar/index.html): From airborne LiDAR of a mountain forest to over a thousand trees: height, crown, stem. Measured: 1,155 trees; 302 per hectare; tallest 38.7 m. - [Stem mapping and tree models (TLS)](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/trees_from_terrestrial/index.html): From a scan made on foot under the canopy to every stem measured, every tree as cylinders. Measured: 23 trees; 548 per hectare; tallest 30.7 m. ## Energy and Utilities Roofs in the sun, trees too close to the wires. - [Roof modelling and solar potential](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/solar_roof/index.html): From open aerial LiDAR of a house to a 3D building and the sun on every roof face. - [Solar potential of a whole street](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/street_solar/index.html): From aerial LiDAR of a street to every roof ranked by its yield, shading included. Measured: 18 buildings ranked; 233 kWp of panels; 225 MWh a year. - [Power line vegetation clearance](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/powerline_risk/index.html): From LiDAR of a power line corridor to wires fitted span by span, tree clearances, work orders. Measured: 19 poles, 19 spans; 36 work orders. ## 3D World Models (AI) Rebuild a room from a video, ask it questions, label it, train on it. - [Ask questions to a point cloud](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/ask_the_cloud/index.html): From a scan of an office to answers measured on the scan, with the evidence. - [Point cloud annotation](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/label_a_scan/index.html): From a classroom scan to a labelled scan, click by click, ready to train on. - [Train a 3D segmentation model](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/train_and_test/index.html): From a labelled scan to a trained model, tested on a room it never saw. Measured: 72.7 % of the new room's points right; mean IoU 53 %; 20 epochs in 2.0 min on a GPU. - [Video to 3D reconstruction](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/video_to_3d/index.html): From a phone video of a room to a 3D scene and every camera, in about a minute. - [Sim-ready digital twin for robot training](https://learngeodata.eu/n3d-demo/v1.69.0-20261020-1921/physics_for_robots/index.html): From a classified office scan to every part with its material, friction and mass, tested in MuJoCo. ## Learn it, or get the software - [3D AI Architect](https://learngeodata.eu/3d-spatial-intelligence-education-mba/): get it. The programme: Neurones 3D and the knowledge to use it to the full. - [Neurones Desktop](https://neurones3d.com/): the desktop engine alone, for any 3D data on your own computer.