Practical Course: Vision Based Navigation

Premeeting

Mateo de Mayo, Jason Chui, Jonathan Eichhorn

Prof. Dr. Daniel Cremers

Version: 01.04.2026

Motivations

DSO

ORB-SLAM

Content of this course

• You can gain practical experience with

  • Visual odometry and localisation / state estimation
  • Vision-based Simultaneous Localization and Mapping (SLAM)
  • Structure from Motion (SfM)

• Implementation of algorithms

• Benefits / drawbacks of specific methods when applied to concrete, relevant problems

• Get familiar with relevant software libraries (Eigen, Ceres, OpenGV, …)

• Learn how to work in teams / on projects

• Improve your presentation skills

Course organisation

• Course takes place during the lecture period

• In Garching CIT in a seminar room TBD

  • Work on your own machine
  • We support Ubuntu 22.04, 24.04, and 26.04 (MacOS and Windows may work with some cmake tweaking)

• Initial phase (first 5 weeks): Lectures & Exercises

  • Mondays 2-4 pm lecture
  • Mondays 4-6 pm exercise session
  • Programming assignments will be handed out every week and checked / graded by the tutors
  • Assignments are worked on individually by every student; each participant should be able to explain their solution
  • Attendance to lecture and exercise sessions voluntary (but highly encouraged)

• Second phase (5-7 weeks): project

  • Work in small groups (1-2 people) on a project
  • Mandatory weekly meeting with tutors to discuss progress and next steps (Mondays 2-6 pm)
  • Implement a specific algorithm / extension / paper, which one tbd
  • Present project outcome in talk and Q&A session (15+5 min per group)
  • Written report on project outcome (10-12 pages, single column, single-spaced lines, 11pt)

Topics covered

Course requirements

• Knowledge of C++ is essential (C/Zig/Rust experience may be accepted)

• Good knowledge of basic mathematics such as linear algebra, calculus, probability theory, and numerics is required

• Prior practical knowledge in robotics and computer vision topics is a plus

• Participation in at least one of the following lectures of the TUM Computer Vision Group

  • Computer Vision I: Variational Methods
  • Computer Vision II: Multiple View Geometry
  • Similar lectures can also be accepted

• Note: If you don’t satisfy the requirements but are still interested in the course, please contact us and we can discuss your case individually

Course registration

• You apply for this course through the matching system: https://matching.in.tum.de/

• Additionally, you have to send us an email:

  • Please specify how you meet the course requirements / if you have attended any related computer vision courses before!
  • Comment on your programming experience in C++! List concrete examples of projects you have worked on or send us your github/gitlab/etc profile.
  • Send all your grade transcripts, in particular showing any lectures on pre-requisite topics (computer vision / robotics / maths) that you have attended to: visnav-ss26@vision.in.tum.de

The deadline for the matching system and prerequisite email is 17.02.2026.

We can only guarantee places to students assigned through the matching process (and fitting the course requirements)!

Watch announcements on the course website: https://cvg.cit.tum.de/teaching/ss2026/visnav_ss2026

The course starts on Monday, 20.04.20267 at 2 pm in CIT seminar room TBD.

Demo

Questions?