3D Vision & Deep Learning
3D Segmentor OS
Build the deep learning stack that turns raw point clouds into classified, queryable 3D scenes. PointNet++, RandLA-Net, and a shipped app. 5 modules.
Note from Florent
Segmentation is where raw scans become actionable. I’ve packed my complete workflow into this OS: from RANSAC to deep learning, you’ll ship a segmentation app that actually runs in production.
What you will build
Module 01
Foundations: 3D Python and object detection
Two full courses are included here. Master the 3D Python stack, then build and train your first 3D object detection engine.
Module 02
Semantic segmentation
Train and deploy deep networks for 3D semantic segmentation. PointNet++, RandLA-Net, real LiDAR.
Module 03
The deployed processing app
Lesson 4 in the curriculum. Build the point cloud processing app you download and run: Python backend, WebGL viewer, and the code and data that go with it.
Module 04
Algorithm forge
Lesson 5 in the curriculum. Region growing, RANSAC, Euclidean clustering, marching cubes, PCA and random forests, scene graph generation, change detection, and large .e57 scans.
Module 05
Unsupervised, labelling and generation
Segment without labels with SegmentAnything 3D, label your own data, and generate 3D assets from text and images.
Your starting resources
Every guide, dataset and code pack below is available to you right now. Nothing is drip-fed.
