3D Vision

3D vision fundamentals for robotics: camera models, multi-view geometry, structure from motion, NeRF, and 3D Gaussian Splatting.

Learning Objectives

1. Camera Models

1.1 Pinhole Model

1.2 Fisheye & Omnidirectional

1.3 Intrinsic & Extrinsic Parameters

2. Multi-View Geometry

2.1 Fundamental & Essential Matrix

2.2 Triangulation

3. Structure from Motion (SfM)

3.1 Feature-Based SfM Pipeline

3.2 COLMAP

4. Visual Odometry

4.1 Monocular VO

4.2 Stereo VO

5. Neural 3D Reconstruction

5.1 NeRF (Neural Radiance Fields)

5.2 3D Gaussian Splatting

6. 3D Vision in Robotics

Exercises

References