CS 395T: Object Recognition

Guidelines for presentations

Shoot for about 20 minutes. Your presentation should provide the class with a clear overview of the paper(s) at hand, but it should also spur our discussion about the significance of the ideas and results. Consider ways in which you can make your presentation as visual as possible. Prepare your delivery; aim to give a presentation that as an audience member you would find informative and easy to follow.

Please think about the following points as you prepare your presentation:

· What is the main problem being solved here, and what is the motivation?

· What do the authors assume in the way they have framed the problem?

· First summarize the technical approach at a high level, and then

· Clarify the most important technical details of the approach.

· What are the main results / experiments performed?

· Can you draw connections to any papers we have already covered?

You may be able to locate the author’s slides for this material online. If so, use your best judgment as to whether you should incorporate them into your own presentation.

Guidelines for demos

To help us gain a more complete intuition about the work we are studying, for each topic one person will present a “demo” of some main idea in a paper we read. When you are in charge of the demo, basically your job is to implement a distilled version of the main technical idea in the paper, and show us some toy examples of how this works in practice.

You might:

· experiment with different types of training/testing data sets

· evaluate its sensitivity to relevant parameter settings

· show (on a small scale) an example in practice that highlights a strength/weakness of the approach

Note that the goal here is NOT to recreate published results. Instead, you are looking to make a small demo that will let us more deeply understand what we have read. Spend some time playing with your implementation, but also put some thought into what would be an instructive toy example to show the class.

For a number of papers, you may be able to find code or binaries provided by the authors online. If so, you should certainly use them, and then spend more time on developing your specific examples or experiments to show the class.