Research Topics in Computer Vision

Hi, I am on lookout for a challenging research topic in Computer Vision which I can work on for about a year and come out with a
successful implementation of the same in MATLAB and then in C. I did some research over the internet and also have come out with few ideas on face ecognition/gesture recognition and 3D modelling and so on. Can you please suggest me a good research topic which you people feel you should have taken it up few years back but didn't actually work on for whatever reason it may be? I am looking out for a topic involving PATTERN RECOGNITION, NEURAL NETWORKS, GEOMETRY AND IMAGE PROCESSING. I know thats a very broad way of naming the topics but I thought I will take ur choices into account before starting off. Waiting for your suggestions on the different topics.
With Regards, Abhishek S
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Abhishek wrote:

Read this guy's articles and you may get some ideas:
http://www.stanford.edu/~dil/invariance /
-- Joe Legris
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Abhishek wrote:

Choose something you are interested in and a project that you think you will use later on in life or at your next job.
I would not suggest you spend any time on Neural Networks. I took some graduate digital signal processing courses in neural networks and they really dont work for anything other than very simple problems that could be solved more predictably by seat-of-your-pants C code.
-howy
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
"howy"

I have to agree in parts with howy. I would study NN only enough to see that you can do the same thing and much more with statistical methods (Monte Carlo, Bayes, Hidden Markov Models, EM, etc). Some people say that NN is nothing more than a statistical method with a neat name.
All this said, I still intend to get back to NN research in the future, but using more realistic neural models.
For computer vision, I would also include in your toolbox things like fourier transforms, wavelets, hough transforms and the basic digital image processing stuff.
Cheers
Padu
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Facial gender discrimination of humans. Note that adrogenous *is* a valid answer. See how well your algorithms stack up against a sampling of college students, using a fixed data set.
Cheerios, Tarkin
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Oh yeah, this reminds me, pick something that involves a lot of photos and measurements of human test subjects. It will be a great way to meet chicks. OK, I got it! Design a computer vision system that detects excessive alcohaul consumption! Imagine the research parties you will have to throw!
-howy

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Abhishek wrote:

Abhishek, I don't know much about what you are trying to do, but I am an engineering undergrad at University of Illinois and read in the school newspaper about a computer science professor who developed software that determined the mood/gender of a person in a poicture. I think this sounds like a good project to work on. Anyway, the newspaper is called the Daily Illini, maybe you could find the article online. Turns out, the monalisa is 60% female-40%male. Ha. Lucas
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.