Seminar Vision and Graphics (Role-Based) (MA-INF 2228, 4CP)

The introductory meeting (for all courses of our group) will be on 10.04.25 14:00. You can get the zoom by registering with paul.roetzer@uni-bonn.de. This seminar will only happen if a minimum of 5 students participate.

Description

This seminar has a different format than most. You will not give a single in-depth presentation about one topic. Instead everyone will read the same paper each week but focus on a different aspect of it. This will require more reading but less depth. As a result of the format, this seminar has mandatory attendance (which will be handled by collecting different roles for the Studienleistung). The concept is based on Role-Playing Paper-Reading Seminars by Alec Jacobson and Colin Raffel but adjusted to adhere to the seminar examination rules in Bonn.

All participants will read one paper from the area of computer vision or graphics every week and analyze it from the point-of-view of a role they are given beforehand. The roles are shuffled every week such that everyone receives every role at least once. During class there is a discussion about the content and wider impact of the paper through the lens of the different roles.

The roles possible roles are:

  • Paper Author: You have written the paper and give a short presentation at the beginning of class with the main contributions. You should also have the paper at hand and be open to answer questions about methodological details during discussion.
  • Scientific Peer Reviewer: The paper has been submitted to a major conference in your field and you have been assigned as the peer reviewer. Complete the full review (review template for an appropriate conference will be provided) and recommend acceptance or rejection of the paper.
  • Archaeologist: This paper was found buried under ground in the desert. You’re an archaeologist who must determine where this paper sits in the context of previous and subsequent work. Find and report on one older paper cited within the current paper that substantially influenced the current paper and one newer paper that cites this current paper.
  • Academic Researcher: You’re a researcher who is working on a new project in this area. Propose an imaginary follow-up project not just based on the current but only possible due to the existence and success of the current paper.
  • Industry Practitioner: You work at a company or organization developing an application or product of your choice (that has not already been suggested in a prior session). Bring a convincing pitch for why you should be paid to implement the method in the paper, and discuss at least one positive and negative impact of this application.
  • Private Investigator: You are a detective who needs to run a background check on one of the paper’s authors. Where have they worked? What did they study? What previous projects might have led to working on this one? What motivated them to work on this project? Feel free to contact the authors, but remember to be courteous, polite, and on-topic.
  • Moderator: You are moderating the discussion about the paper. You should make sure that every role gets the chance to present their point of view and prepare at least three interesting discussion points.

Author and Peer Reviewer are the core roles of every week. The distribution of other roles might vary depending on the paper and number of participants.

Paper List

As-Rigid-As-Possible Surface Modeling

Paper: Sorkine et al. SGP 2007.

ImageNet Classification with Deep Convolutional Neural Networks

Paper: Krizhevsky et al. NeurIPS 2012.

NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis

Paper: Mildenhall et al. ECCV 2020.

Functional Maps: A Flexible Representation of Maps Between Shapes

Paper: Ovsjanikov et al. SIGGRAPH 2012.

VectorAdam for Rotation Equivariant Geometry Optimization

Paper: Ling et al. NeurIPS 2022.

Object Recognition from Local Scale-Invariant Features

Paper: Lowe. ICCV 1999.

Neural Scene Graphs for Dynamic Scenes

Paper: Ost et al. ECCV 2021.

MeshGPT: Generating Triangle Meshes with Decoder-Only Transformers

Paper: Siddiqui et al. CVPR 2024.

Requirements

  • There are no formal course requirements.
  • Prior knowledge in the area of computer vision or graphics through lectures in these topics is recommended.

Participation

You can register for the seminar until April 13th by writing an email to laehner at uni-bonn dot de with your study program and matriculation number. The course has capacity for 5-10 students. If there are more registrations than slots, they will be given out randomly. You will get an answer on April 14th. Topics and roles will be distributed on eCampus.

The seminar will happen on Mondays 10-12 in 3.035b.

Grading: paper author and peer reviewer are the main roles which will be graded. Additional roles will be used for the Studienleistung and bonus points. (Exact grading scheme will be in eCampus.)

Attendance: The Studienleistung will be received by participating in each type of assigned role in-presence at least once. You can switch roles with other students to make duplicate roles fit your schedule and you can have role re-assigned to fulfill the Studienleistung if you have a doctor’s certificate for the missed role.