1
$\begingroup$

How to constrain the object detection problem?

It would be tempting to ask for precision such as: "It depends, does it have to be viewpoint invariant, is there going to be viewpoint changes, etc.?" Actually, it is exactly what I want to know. I have a specific application in mind and I need a structured way to help me constrain and define correctly the problem. I am pursuing a graduate degree but I am new to the field.

I mean, at first I read the literature about things such as binary keypoint descriptor, etc. but in the end, I am putting the cart before the horse, not having constrained my problem correctly, I cannot know or prove whether binary keypoint descriptors or else are a good starting point for my application...

Unfortunately my adviser is unavailable for a week or two and my colleagues are not specialized in that field.

$\endgroup$
2
$\begingroup$

A long time ago I have come across a very good advice: research should be problem-oriented, not approach oriented. That means find a problem you want to solve, and then look for appropriate methods. Not the other way around. So, I would recommend starting with a specific problem, such as a real-world application.

For example, maybe you want to build a system with a camera looking at your bookshelf, which detects a particular book. Or maybe you want to build a mobile robot with a camera, which can follow you around. Or a video player program, which detects whether you are looking at your computer screen, and pauses the video when you look away (some cell phones do that now). Or a program that can organize your photo collection by detecting and recognizing people, and maybe landmarks, like the Eiffel Tower. The possibilities are endless.

So the algorithm is as follows:

  1. Find a real-world vision problem that you find interesting.
  2. Make as many restrictive assumptions as you can to scope it down and still keep it interesting.
  3. Start reading papers to find applicable algorithms.
  4. Understand the limitations of existing algorithms, and come up with your own ideas to address them.
$\endgroup$
  • 1
    $\begingroup$ I think your first paragraph should be taught systematically in every graduate curriculum ;) $\endgroup$ – Doombot Sep 24 '14 at 13:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.