The topic of your research review is free to choose, but it must be Computer Vision related. Focus on a Computer-Vision algorithm and understand its details. Look at the intro slides for some inspiration!
The deliverables for the research are the announcements of your idea and your team, one presentation, and a chapter in the course report.
For presentations you will have a maximum of 10 minutes for your talk plus 5 minutes for questions. Your goal is to present something interesting to your fellow students, so they learn something from your talk. It is okay if only one team member presents.
Announce your team and the topic you want to work on via Elearning. If you submit early and want early feedback, please notify me via mail. In the announcement you should be able to answer the following questions:
- Who is on your team (names of one or two student(s) )?
- What is the problem/question/topic that you will be reviewing?
- Did you find existing and ecxiting literature, already?
I will notify you if the project topic is okay and might give you some hints. Note that for the announcement you already need to look at some related work, which will help you for the next part.
The objective for your presentation is to carry out a bibliographic study and present the result to the class. Choose some papers (not less than two, preferably not more than five) that are closely related to your topic. Papers should be interesting and not too hard to follow. The lists (provided below) might be a starting point to initiating the study and are by no means a constraint. You will also find papers through citations. Study the papers in depth. Find how they are connected. Pay attention to related work and, through citations, try to identify papers that are more relevant, more interesting, or more recent.
For the related work presentation, you will be evaluated based on your choice of papers, the way you have connected them into a story, and how you present them. You should focus on the main ideas. Implementation details are typically irrelevant. During the presentation you should answer the following major questions:
- What are the most relevant readings / main ideas? (2-5 papers)
- What are the existing methods?
Summarize your literature research (Part 1) as a chapter in the course report. The report also contains the solution to your homework in additional chapters. Reports should be submitted by each team member. If there is an unequal distribution of contributions within your team describe it in the report (e.g., student A did 60% of the work while B only did 40%), and I will consider it for grading.
Keep the sections in your report short and to the point, especially the homework descriptions. For example, a small result described and evaluated well will earn more credit than a long text where no aspect was done well. Be accurate in describing your solutions. Make sure to add references to all related work you reviewed or used (especially in the related work chapter). Please submit any supplementary material that is important to evaluate your work (e.g., code, images, videos), but make your report as self-containing as possible. It should be possible to understand your solutions from the report only.
Below you will find some inspirations for your research:
Possible Topics
A list of possible topics for the CV course. Furthermore, look at the intro slides for some inspiration!
- arXiv.org CVPR list: https://arxiv.org/list/cs.CV/recent
- CVF open access: http://openaccess.thecvf.com/
- NeurIPS: https://papers.nips.cc/
- ICLR: https://openreview.net/group?id=ICLR.cc
- A paper list of Deep-Learning related techniques organized into categories and sub-categories (from University of Rennes): https://sif-dlv.github.io/oral/
- Computer Vision companies and products (not up to date anymore): https://www.cs.ubc.ca/~lowe/vision.html
- Resources, such as image databases and books (from MIT): http://6.869.csail.mit.edu/fa19/materials.html
- A list of project ideas (from MIT): http://6.869.csail.mit.edu/fa19/project.html
- YouTube channels that discuss recent science breakthroughs (mostly deep learning, such as Yannic Kilcher or tow minute papers )