Lectures

Computer Vision

What a vehicle, a robot or a camera system has to do to turn pixels into an understanding of a scene. The course covers the classical foundations and the learned methods that replaced parts of them, and shows where each still has its place.

What you learn

  • Image-based feature extraction and matching, and what makes a feature stable.
  • Segmentation: separating what belongs together in an image.
  • Motion estimation and tracking over time.
  • 3D reconstruction from several views, and what geometry can tell you without learning.
  • Classification and detection with neural networks.
  • Processing point clouds from LiDAR sensors.

How it is taught

Every unit is followed by a practical application example from industry, so the methods are used on data that actually behaves badly. The lab’s research vehicle provides some of the recordings.

Prerequisites

A working knowledge of machine learning (the Bachelor course covers it) and Python.