Deliverables and grading policy
You will have the following deliverables through the course:
Assignments (25% total, number TBD)
The assignments have to be done individually. However, you can discuss with each other.
Submit a text file along with the assignment naming those you discuss with.
Paper summaries (10%)
Being able to read a paper and grasp both what the paper is about, and how it falls short is a crucial research skill no matter what area you work in.
The course schedule lists some readings for certain lectures.
Before the lecture, you are required to read the paper and answer a few discussion questions about each paper on CMS.
In the lecture, we will discuss these discussion questions in groups (if possible).
The biggest requirement of the course is a final research project.
There are two requirements for this project:
To understand the scope of this project, here are a few hypothetical example projects:
- It must be a research project. This means that it should create new knowledge: it should tell the broader community something that they did not know before, at the level of a workshop paper. Note that it need not be research in computer vision; instead, it might simply use established computer vision techniques to answer a research question in your field of study. However, it should be a research question: a question to which the community does not know the answer.
Finding such problems is usually the first step of a PhD.
- It should involve computer vision in some manner. Use your judgement. This is a computer vision course after all.
- Train an object recognition system for a new, or rarely tested domain, e.g., satellite images
- Evaluate the effect of the choice of activation function on neural network training on multiple datasets
- Evaluate the accuracy of state-of-the-art stereo techniques for very wide baselines
- Use simple computer vision techniques to count microbes in cells and use this to answer research questions about disease spread
PhD students are encouraged to do projects in their research area. You can use the same project for multiple classes provided it is of a correspondingly large scope.
However, if you are working with me on research, you must carve out a research project that will not be advised by me.
The project itself will have the following deliverables:
- A project proposal. The parameters of this will be made clear soon
- A project check-in, with an updated project proposal if necessary. This is to allow for changes in teams, new directions based on preliminary results etc.
- An intermediate project report, which will be reviewed by your peers.
- A final project ``spotlight'' presentation (4 minutes)
- A final project report.
The project should be done in groups of 2-4. Please try to find a partner: research can be lonely :)
Peer review (15%)
The final component of this course will be a peer review. You will be asked to review your peer's projects. We will aim for upto 3 reviews for each project. Be constructive and informative, you will be evaluated on the quality of this!
We will share examples of good reviews as a suggestion.