3D Geodata Academy

Article materials

The code and data behind every article

Every tutorial I publish ships with a script that runs and the scene it was written against. The code is public on GitHub. The datasets are one email away, because they are too big to paste out of an article.

The kits

9 kits, newest first. Each one reproduces every number in its article.

An office floor point cloud coloured by walking distance to the nearest exit, with escape routes 2026-10-29

Walking distance to the nearest exit, from a scan

One multi-source Dijkstra from the exits over a 16-neighbour grid, routes by descent, dead-end corridors from graph bridges, and a what-if for every blocked door. CPU only.

A classified indoor scan with its floor coloured by wheelchair reach and every issue numbered 2026-10-22

Wheelchair accessibility audit from a 3D scan

Clear width by distance transform, turning space by morphological opening, reach by connected components, then a numbered issue list. CPU only.

Three IFC wall solids, each labelled with its measured thickness, standing among the wall points of an S3DIS office 2026-10-14

Wall thickness from a scan, written to IFC

Find the two faces of a wall as density peaks, pair them, measure the gap cell by cell, write an IfcWall and prove it. CPU only.

Four vector walls stood back up over a classified classroom point cloud, one in orange 2026-10-05

Scan to floor plan, from wall pixels to a DXF

Turn the wall points of a classified room scan into clean vector walls and a layered DXF any CAD tool opens. CPU only.

A Gaussian splat of a forest cabin rebuilt as a clean textured mesh 2026-09-22

Gaussian splat to textured mesh

Filter, densify and Poisson-mesh a trained 3D Gaussian splat, then bake a real texture. CPU only.

A LiDAR point cloud processed in Python, from raw tile to terrain model 2026-07-14

LiDAR ground filtering to a terrain model

Read, thin and index a real LiDAR tile, filter the ground, grade it against the survey, write a DTM. CPU only.

A living-room point cloud labeled into walls, floor and furniture 2026-06-22

Open-vocabulary 3D semantics on a real room scan

Lift 2D labels onto a real room scan, fuse them across cameras, vote out the flicker, slice a plan. CPU only; model outputs simulated.

A 3D point cloud of a toy set labeled by painting brush strokes on photos 2026-03-23

Label a 3D scene by painting on photos

Paint labels on photos with an OpenCV brush, project them to 3D through the cameras, spread them with a KD-tree, export PLY and GLB. CPU only.

A toy construction set rebuilt as a 3D point cloud from nine phone photos 2026-02-11

3D models from photos with Depth-Anything-3

Depth, camera poses, a registered point cloud, ground and object segmentation, a voxel mesh and a GLB, from nine phone photos.

All of it in one place: github.com/florentPoux/3d-data-materials

Cover of 3D Data Science with Python by Dr. Florent Poux
Go deeper

3D Data Science with Python

Building Accurate Digital Environments with 3D Point Cloud Workflows

Every kit here is one pipeline. The book is the discipline underneath them: 690 pages from the Python foundations through point cloud processing, meshing and the engineering that keeps a 3D workflow standing up in production. If these pipelines make sense and you want the structure they came from, that is where it lives.

O’Reilly Media, 2025 · ISBN 9781098161330 · 15,000+ readers

See the book or buy it on O’Reilly

Where the articles live

The kits are the endings. These are the articles they end.

If you want the path, not the pieces

A kit teaches one pipeline. These teach the discipline.

Dr. Florent Poux

Dr. Florent Poux

Founder and Lead Instructor at the 3D Geodata Academy. 15+ years on the automation of reality capture, from point clouds and photogrammetry to spatial AI, and 1,700+ citations across 60+ peer-reviewed publications.

Each kit page hands you the code, the scene and the finished output. The code is public on GitHub under the MIT license.

Scroll to Top