Computer Vision

Prof. Thomas Brox

Computer vision is a very active research field with many practical applications, for instance in quality control, robotics, driver assistance systems, and many more. The ultimate goal of computer vision is to imitate the great capabilies of the human visual system, allowing the computer not only to record images but also to interpret them. Significant progress has been made in recent years. This course introduces the most important concepts in today's Computer Vision research. The exercises will consist of programming assignments in Python, where you will learn to implement and/or test some of the most important techniques presented in the lectures. Lecture: (2 SWS) Monday, 10:15-11:45, Room: 101-01-009/13 Exercises: (2 SWS) Monday, 16:00 (first session Oct. 21), Room: 101-01-009/13 Contact persons: Karim Farid, Simon Ging Please ask questions in the exercise sessions or via the ILIAS Forum instead of contacting us by email. Please register for the exercises using the TF pool access form at the bottom of this page. Beginning: Monday, October 14, 2024 ECTS Credits: 6 Requirements: Fundamental mathematical knowledge and programming skills in Python. Image Processing and Computer Graphics teaches important technical knowledge that won't be repeated in Computer Vision. Thus, prior attendance of Image Processing is highly recommended. Language: The lectures are given in English. Exam: Written exam.

---

Material:

Lecture material will be updated during the course. Recording files are about 200MB each.

Class 1 (14.10.):	Introduction	Recordings
Class 2 (21.10.)*:	Diffusion filters, TV minimization	Recordings
Class 3 (28.10.):	Spectral clustering	Recordings
Class 4 (4.11.):	Deep learning	Recordings
Class 5 (11.11.):	Object detection and semantic segmentation	Recordings
Class 6 (18.11.):	Optical Flow	Recordings
Class 7 (25.11.):	Deep learning based optical flow	Recordings
Class 8 (optional, no actual class):	Video segmentation	Recordings
Class 9 (2.12.):	3D geometry and camera calibration	Recordings
Class 10 (9.12.):	Stereo reconstruction	Recordings
Class 11 (16.12.):	Disparity estimation and scene flow	Recordings
Class 12 (13.1.):	Structure from motion	Recordings
Class 13 (20.1.):	Deep learning based 3D vision	Recordings
Class 14 (27.1.):	Image generation	Recordings
Class 15 (3.2.):	Next generation deep learning	Recordings

---

Exercise material:

Short introduction to git
Setup instructions
Exercises on github
TF pool access form

NumPy quickstart
PyTorch cheat sheet
PyTorch for NumPy users

---