Login
Computer Vision II: Multiple View Geometry (IN2228)
Offered in Hybrid Format in SS 2022, TU München
Registration
If you plan to attend, please register for the course in TUMonline.
Later during the semester, you will have to register for the exam.
Lecture
All the lectures and exercise classes will be recorded and streamed automatically. Watch them Live on RGB by logging in with your TUM account.
Time and Location:
Wednesday 12:15 - 13:45, "Interims I" Hörsaal 2 (5620.01.102)
Thursday 11:00 - 11:45, MI Hörsaal 1 (5602.EG.001)
Lecturer: Prof. Dr. Daniel Cremers
Start: Thursday, 28.04.2022
The lecture is held in English.
Exercises
Location: "Interims I" Hörsaal 2 (5620.01.102)
Time: Wednesday, 16:00 - 18:15
Organization: Simon Weber, Tarun Yenamandra
(If you have any questions please email to: mvg-ss22@vision.in.tum.de)
Start: Wednesday, 04.05.2022
Exercises are split into theoretical and practical parts.
MATLAB
The programming exercises are done in MATLAB. You don't need extensive prior MATLAB knowledge. If you have never used MATLAB before, we recommend following a basic tutorial. Please install MATLAB on your laptop before the first exercise using the university's student licenses.
- MATLAB cheatsheet: http://web.mit.edu/18.06/www/Spring09/matlab-cheatsheet.pdf
- MATLAB on your own computer: http://matlab.rbg.tum.de
- We don't use much beyond core MATLAB functionality. You may need the Image Processing Toolbox for reading and displaying images.
Note on MATLAB version:
We suggest you to use a MATLAB version 2016b or later. For the solutions we make use of function definitions inside Matlab script files. This feature was added only in 2016b. So if you have an earlier version, we suggest you to upgrade (alternatively you can move all function definitions to separate files). See also https://de.mathworks.com/help/matlab/matlab_prog/local-functions-in-scripts.html.
Summary
The lecture introduces the basic concepts of image formation - perspective projection and camera motion. The goal is to reconstruct the three-dimensional world and the camera motion from multiple images. To this end, one determines correspondences between points in various images and respective constraints that allow computing motion and 3D structure. A particular emphasis of the lecture is on mathematical descriptions of rigid body motion and perspective projection. For estimating camera motion and 3D geometry we will make use of both spectral methods and methods of nonlinear optimization.
Lecture Material
Slides and exercises can be downloaded here.
Lecture Videos
You can find recordings of a previous iteration of this course on Youtube.
Literature
Yi Ma, Stefano Soatto, Jana Kosecka, Shankar S. Sastry. An Invitation to 3-D Vision (use your TUM-ID to login to Springer Link)
