Computer Vision 1

Carsten RotherDmitrij Schlesinger, Holger Heidrich, Winter semester 2016/2017

Computer Vision is a science that develops models and methods for understanding, analysing, acquiring and processing images, and more generally high-dimensional “visual” data. Computer Vision is a discipline which makes use of many other fields such as discrete optimization, machine learning, human-computer interaction and computer graphics. We offer two courses in Computer Vision: Computer Vision 1 runs every winter semester, and computer Vision 2 every summer semester. Computer Vision 1 considers predominantly the “physical aspects” of computer vision, such as geometry, reconstruction, object tracking, and image processing. In particular we cover in detail how to obtain a 3D reconstruction from a set of 2D images. Computer Vision 2 will look more at the semantic aspects of computer vision, such to recognize and segment all objects present in an image. Both courses focus on algorithms, modelling and applications. In contrast to this the courses in machine learning focus more on theoretical aspects of inference and learning from data.

ModulesINF-LE-MA INF-B-510 INF-B-530 INF-BAS2 INF-BAS7 INF-VERT2 INF-VERT7 INF-VMI-8 INF-PM-ANW INF-PM-FOR MATH-MA-INFGDV

Lectures: Friday, 2. DS, 09:20 – 10:50 Uhr, INF E023, Start: 21. October 2016.

Practice: Tuesday, 2. DS, 9:20 – 10:50 Uhr, room: E046 (changed!) or (you choose) Wednesday, 3. DS, 11:10 – 12:40 Uhr, room APB E067 (changed!), Start: 11. October 2016.

Prerequisites: good knowledge of maths (linear algebra, optimization), programming (C++).

Credits: 2/2/0, oral exam,

Enrollment: jExam,

Attendees: max. 60.

Note: lectures are held in German with slides in English. There are two course books. The first one is: “Computer Vision: Algorithms and Applications” by Richard Szeliski which can also be found online: http://szeliski.org/Book; the second one is: Multi-View Geometry by Hartley and Zisserman, Cambridge Press 2004. This course is a prerequisite for the course in SS‘17 “Computer Vision II: Models, Inference, and Learning”.


Info about the oral exam: The main part of the exam is about the lectures. It is possible that you will also be asked about the exercise; if you got at least 8 (out of 13) points the questions that regard the exercise will concentrate around what you did, otherwise they will cover the whole set of exercise tasks.


News:

07.03.2017: The updated points for exercises are here (07.03). Check it please and contact me in case of questions (mistakes, inconsistencies whatever).

07.01.2016: Reminder: the deadline for the third exercise “Panorama Stitching” is 11.01 !!! Ex.4 “Tracking” starts on 17./18.01.

04.01.2016: The page seems to work again. Please check your points (below).

23.12.2016: Due to some technical reasons I can not upload the actual “points”-file. Hope, it will work soon.

04.12.2016: Ex.3 “Panorama stitching” starts on 06./07.12.2016. There will be no defenses for Ex.3 — only submissions per email. So we continue defenses for Ex.2 on 13./14.21.2016.

30.11.2016: Reminder !!! The deadline for submitting the second exercise “Filtering” is today 30.11 !!! (per e-mail, no matter whether you defended it or not). For those who did not defend so far, there will be additional time slots announced later.

14.11.2016: There will be two further time slots for defending “Filtering” exercise — on 29.11 and 30.11, i.e. the deadline is extended to 30.11 (strict). No exercises on 22.11 and 23.11 (of course, you can come and work, but I am not in).

10.11.2016: Information: Am 30.11 (14.00 in 2024 oder 2026) kommt eine Besucher (Vincent Lepetit) der über einen lernbaren SIFT descriptor (LIFT) redet. Mehr Infos unter: http://cvlab-dresden.de/computer-vision-talk-series/

10.10.2016: Could not get a place in jExam? Don’t worry: the only reason is, that the number of computers is limited in the class room. So come to the exercise with your laptop. (And preinstall OpenCV in debug mode.) Please enroll in jExam as waiting or for the lecture in that case, so that we have your contact data.

 


Scripts:
Lectures: (slides available around time of lecture)

14.10. : no lecture in the 1st week (but exercise)
21.10. : Introduction
28.10. : Image processing 1
04.11. : Image processing 2
11.11. : Geometry
18.11. – no lecture!
25.11. : Geometry
02.12. : Geometry
09.12. : Geometry
16.12. : 3D reconstruction and Decision Trees
08.01. : Tracking and Detection
13.01. : Tracking and Detection
20.01. : Tracking and Detection
27.01. : Neural networks and End-to-End Pipelines
03.02. : Neural networks and End-to-End Pipelines

Practice:

There will be 4 topics for exercises. Each will have different tasks from which you gain 1 to 4 points. You need 8 ot of 13 points to pass the CV1 exercise course and at least one point from each of the 4 exercises. The  first exercise gives you 1 point, the other up to 4 points each.

Exercise 1: 11./12.10.: Set up your OpenCV environment and program a simple image manipulation, Slides, QtCreator Project File,  | sln_comments, solution proposal, current Points you have

Exercise 2: 25./26.10.: Filtering techniques

Exercise 3: 06./07.12. Panorama Stitching, Data.

Exercise 4: 17./18.01. Tracking, Data and Code. IMPORTANT: To hand in your results, please come to the sessions. You can show and explain the code to me during the session when ever you are ready. Do not email your code to me.