Learn to See the World in 3D - From Point Clouds to Learned Representations
In a world increasingly shaped by 3D data, from autonomous vehicles and architectural design to AR/VR experiences, point cloud has become a powerful solution for men and machines to visualize and understand spatial information. This talk takes you on a journey from raw points to perception: exploring how individual 3D points in point cloud come together in space to form meaningful structures. We’ll start with the fundamentals of point cloud and 3D representation, then dive into how neural networks learn to recognize, reconstruct, and even compress the data. The session combines theory with hands-on experience, participants will learn how to visualize point clouds, build a simple deep learning model, and observe how machines learn to solve some computer vision tasks.
- Date: December 3, 2025
- Time: Theory: 14:00 - 14:50 / Coffee 14:50 - 15:10 / Lab 15:10 - 16:00
- Location: Nancy-Salle A008 Jean Legras
- Instructor: Mohammedreza Ghafari, Ph.D. Candidate