INP.32888UF+INP.32989UF Robot vision (2 VO+1 KU, SS)

Lecturer: Friedrich Fraundorfer (fraundorfernoSpam@icg.tugraz.at)

Learn about 3D reconstruction from images as well as computer vision methods needed for mobile robotics.

Important information for SS 2022

All the lectures are planned to be held in presence. The lectures will be held in i11 and the lecture room should be large enough for all attendees. The lectures will be streamed using Tube. In addition the lecture topics are available as pre-recorded videos.

The practical is organized as programming assignments which can be carried out from home. The practical is organized as group work in groups of 2. This is important such that you do have the possibility to discuss the topics of the practical with a peer. Assignment handouts will be organized as hybrid events. There is limited seating for attending in presence (pre-scheduling using the Teach Center) but the events are held in a hybrid form using Webex. The access information can be found in the Teach Center for registered students. In addition there will be assignment interviews with the individual groups in presence and Q&A sessions with the tutor in hybrid form. A schedule for these events will be put on Teach Center.

Course description

The course is organized as a lecture and practical. It is advised to take both of them at the same time. To pass the lecture an exam has to be taken. Exam dates will be published in TUG-Online. The practical part is organized as programming assignments and carried out in groups of 2.

Lecture topics

  • Projective geometry
  • Image formation and camera calibration
  • Geometric algorithms (Fundamental matrix, Essential Matrix, Triangulation)
  • Robust estimation (Ransac)
  • Features and matching
  • SfM
  • Bundle adjustment
  • Stereo matching
  • Multi-view stereo
  • Deep learning for depth estimation
  • Depth cameras

Practical part (INP.32989UF)

The practical consists of programming assignments. The practical will be done as group work in groups of 2 students. There will be 3 assignments to be completed through the semester. The assignments need to be programmed in C/C++ with the help of the OpenCV computer vision library. The first assignment will be about camera calibration. The second assignment will be about feature matching and 2 view geometry estimation. The third assignment will be depth estimation using deep learning.

A tutor will assist you and be available for Q&A about the assignments.

Assignment schedule:

  • Assignment 1
    • Handout: 2.3.2022 (hybrid event)
    • Deadline: 29.4.2022
    • Interviews: 4.5.2022/18.5.2022 (individual slots for groups)
  • Assignment 2
    • Handout: 23.3.2022 (hybrid event)
    • Deadline: 17.5.2022
    • Interviews: 25.5.2022/1.6.2022/15.6.2022 (individual slots for groups)
  • Assignment 3
    • Handout: 6.4.2022 (hybrid event)
    • Deadline: 14.6.2022
    • Interviews: 22.6.2022/29.6.2022 (individual slots for groups)

  • Q&A sessions: 9.3.2022, 16.3.2022, 30.3.2022, 27.4.2022, 11.5.2022, 8.6.2022

Grading

Grades for the lecture can be obtained by taking a written exam. Exam dates will be published in TUG-Online. The grades for the practical (INP.32989UF) are independent of the lecture and will be determined based on the submitted programming assignments.

Times and dates:

  • First lecture will be held in presence in i11 on the 1.3.2022 from 14:30-16:00 (Streaming will be available via Tube for registered participants).
  • First event for the practical (RV KU) will be on 2.3.2022 as an hybrid event. The topic will be an introduction to the organization of the practical and the handout of the first assignment sheet.
  • The main exam date will be at the time of the last lecture slot 28.6.2022 from 14:30-16:00 as a written exam.

Lecture slides: 

(slides will gradually appear here)

 

Lecture slides 2021:


  •  

News: 

(website updated on 17.02.2022)