CS 598 LAZ: Computer Vision: What Will Stand the Test of Time?

Meeting time: W F 11AM - 12:15PM, 1320 DCL

Instructor: Svetlana Lazebnik  (slazebni -at- illinois.edu)

TA: Aiyu Cui (aiyucui2)

Contents of webpage: reading list, requirements, schedule

Announcements, discussion: Campuswire

Submission of deliverables, grades: Canvas

Overview

This is an advanced graduate seminar attempting to address meta-level issues in computer vision research (and in AI more generally). The format will consist of a few introductory lectures by the instructor, followed by student-led presentations in groups of two to four. Instead of focusing on cutting-edge technical topics, as is typical, this seminar will take a "big picture" view of the research landscape and answer questions including, but not limited to, the following. What "classical" papers from the history of the field are still relevant today? What will count as a "classic" several decades from now? What ideas and practices from other fields (neuroscience, cognitive science, social science, etc.) should inform our research? How are breakthrough technologies such as generative models and large language models likely to affect society, and how should their development be directed or regulated? What will be the roles of academic vs. industry researchers going forward? Etc. Apart from the group presentation, requirements will include a final project or paper, and classroom participation. Interest in active in-person discussion is a must!

Detailed Topic List and Readings

Requirements

Participation

Lively discussion is key to this class. You are expected to be present most days and participate in discussions both during class and on the Campuswire discussion board (a thread for questions and comments will be created for each lecture).

Group presentation

Students will form groups of two to four, jointly choose a topic you're fascinated by, and develop a presentation about it. Topics can be derived from the reading list, but are not limited to it. Presentations are to be done in person, and each group member must deliver a portion of the presentation.
  • Possible types of presentations include (but are not limited to):

    • Historical overview or survey of a specific sub-field or research area

    • "Deep dive" into the work of a specific thinker or researcher (or small number of researchers)

    • "Deep dive" into a famous book or paper (or small group of papers)

    • Overview of a contentious issue or question with multiple viewpoints represented

    • Polemical presentation -- argue a specific point of view, such as "Peer review should be abolished" or "Benchmarks are holding back computer vision research"

    • Debate -- for contentious questions, this would be a particularly fun format to try. The debates should roughly follow the Oxford format.
      • The debate must be centered on a debate question that admits a yes/no answer. Here are some possible debate questions (which should be fleshed out with necessary definitions/justifications/clarifications):
        • Is embodiment necessary for development of AGI?
        • Is "Strong AI" true?
        • Is reaching AGI just a matter of scaling up, or are there important fundamental problems that are still unsolved?
        • The "bitter lesson": True or not?
        • Does AI pose a significant existential risk to humanity?
        • Will AI cause mass unemployment in the future?
        • Can AI be conscious?
        • Can AI be creative?
        • Is academic AI research doomed to fall behind industry?
        • Publishing models: should peer review be abandoned?
        • Will open-source development win the day?
        • Does AI/ML have an overpublication crisis?
        • Does AI/ML have a replication crisis?
        • Are benchmarks starting to hold back progress in AI?
        • Is a policy framework like data dignity necessary to ensure ethical use of data by AI applications?

      • The affirmative team (or person) makes arguments that support a "yes" response, and the negative team makes arguments that support a "no" response. To argue on a given side, the position doesn't have to match your beliefs exactly, but you have to be willing to make the strongest possible arguments with conviction. All four participants must agree on their debate question at the time of signup and finalize the wording of the question by the submission of the reading lists. Apart from agreeing on the question, each team prepares and submits their materials separately.
      • The debate starts with 15-minute opening statements by the "yes" and "no" teams. Statements should feature prepared slides.
      • Following the opening statements, each team will briefly confer and deliver a 5-minute rebuttal to the other's opening statement.
      • Following the rebuttal, there will be approximately 20 minutes of audience participation, where members of the audience will ask questions directed to any of the teams. Lana will moderate and direct who should answer the questions.
      • At the end, the audience will vote on which side, affirmative or negative, made the more convincing arguments.


  • Signup: Shortly after the first lecture, a link to a signup sheet will be shared with all the registered students. Signup deadline for registered students is Monday, August 28. If you fail to sign up on time, you will lose 10% of your presentation grade, and may not receive a presentation spot at all, making it impossible to pass the course.

    After the deadline for registered students, any unregistered students interested in taking the course are free to sign up for the remaining spots. The presentation schedule will be finalized by Friday, September 1.

    Keep in mind that the act of signing up for a topic is a commitment to your teammates, the instructor, and all the other students in the class. Therefore, if you are unsure whether you will stay in the course, we urge you to make this decision now if at all possible.

  • Outline and reading list (10% of presentation grade): By the end of Sunday the week before you are scheduled to present (i.e., either 10 or 12 days before your presentation), one member of the group must email Lana a brief outline and list of references you plan to cover. In case of a debate, instead of an outline, you should submit the exact wording of your debate question or statement. This is a hard deadline -- failure to submit the outline on time will forfeit this portion of the grade for your team and may negatively affect the ability to schedule a practice presentation.

  • Practice presentation (20% of presentation grade): After you submit your outline, Lana will schedule a time for your practice presentation approximately a week before your presentation date. The goal is to enable feedback to ensure the highest possible quality of the in-class presentation. All group members must attend. The practice presentation is not expected to be polished or 100% complete, but the grading will be based primarily on evidence that the group is taking the preparation seriously. Even before the "official" practice stage, you are highly encouraged to consult with Lana or Aiyu during office hours about your reading list or draft slides.

  • Slides and final reading list (30% of presentation grade): By the end of the day before the scheduled presentation, the group must email Lana the link to your slides. Acceptable formats include PowerPoint, PDF, and Google Slides. Be sure to put your names on the title slide and a final reading list, with links, on the final slide. You can still make changes right up until your presentation, and the link will not be available to the students until after the class, but late submission of the link will forfeit this portion of the grade. Please see the note on credit attribution below, as failure to follow those guidelines will negatively affect this portion of your grade and may even be considered an academic integrity violation.

  • In-class presentation (40% of presentation grade): The presentations will be graded based on clarity, technical depth and/or educational value, successful synthesis of content from multiple sources, ability to involve the audience, and responsiveness to feedback from the practice presentation.

Barring truly exceptional circumstances, all group members will receive the same score for all of the above components.

Guidelines for creating a successful presentation:

  • The initial reading lists are not meant to be binding. You should feel free to modify them based on your subsequent preparation. Your goal is to identify a thematically coherent group of readings that will have high educational value and stimulate discussion. You are not expected to cover every part of every source, and your presentation typically should not be organized as a sequence of single-paper summaries. Make sure your list is neither too broad nor too narrow in scope.

  • Think of yourselves as professors for a day. If you are giving a technical lecture or survey, you should strive to be comprehensive and understandable, to expose the core technical ideas without going into excessive or overwhelming detail. If you are giving a less technical or more polemical talk, you should still strive to be as clear and informative as possible, to generate active participation, and to lead the class in a constructive discussion.

  • When preparing your slides, and especially when merging slides provided by different group members, pay particular attention to integration. Make sure that slides covering different aspects have a consistent organization, level of detail, and notation. Think about the overall "story" you will tell as you are going from one aspect of the topic to the next.

  • Be sure to place your topic in the context of the entire course. Whenever appropriate, take care to point out any specific connections to other presentations that came before.

  • Wherever possible, you should bring a critical perspective to your topic. Go beyond simply describing the techniques. Compare and contrast different papers, question assumptions, expose possible flaws and limitations, suggest alternatives and/or directions for future research. If you're talking about old techniques, explain why they are relevant today and how they have influenced subsequent work -- or how they have proven to be misguided. On the other hand, if you are covering material that is not peer reviewed, you should maintain skepticism about any confusing statements or extraordinary/unsubstantiated claims.

  • Wherever possible or appropriate, you are highly encouraged to include a demo component. You can use code or demos from the Web or implement your own (this could even coincide with your project if your presentation comes later in the semester).

  • Be sure to involve the class. When you are developing your presentation, identify places where you can ask other students for input, or topics that you want to open up for discussion.

  • Because timing is hard to predict, you need to maintain some flexibility in terms of the topics you will cover. It is a good idea to have one or two sections in the latter half of your slides that you can skip depending on the time. When you are presenting, keep an eye on the time and adjust the pacing towards the end accordingly.

  • Use of external sources and credit attribution: Be sure to explicitly give credit whenever you use material from other sources. If you "borrow" any slides or graphics, be sure to link to the original source at the bottom of each slide. If you show a demo based on somebody's code, be sure to clearly announce this. Failure to follow these guidelines will hurt your score for the slides, and may even be considered an academic integrity violation. It is not acceptable to use an entire slide deck from another source "as is" as the basis for your presentation.

Group project

You are encouraged to work on the project with your presentation group and tie it to the topic of your presentation. However, it is also fine if your project is unrelated to your presentation and your project group is different from the presentation one (groups should not be larger than four). The project may take the following forms: Project deliverables: