Open source class scheduler

I was just wondering how useful would it be to have an open source class scheduler. I mean, create a GTK or Qt based scheduler using Hungarian Matching or some other Operational Research algorithm in order to schedule teachers and classes or other types of tasks.
I've seen commercial versions of around U$100.00, but I'm still not sure how much benefit it would be to the open source community.
I'm tracking this post's visits, so, the number of Google searches that lead here will be tracked and I'll decide if it's worth based on this number and comments to this post.
Thanks for your feedback

GPUWire - A GPU implementation of the livewire image segmentation algorithm

This video shows one of the results of my master thesis which is a solution to the Single Source Shortest Path problem through the use of a GPU.
The idea belongs to the field of GPGPU and my approach was using the GPU in order to parallely expand vertices just like the Delta-stepping algorithm.
Source code, as well as my master thesis can be found at http://code.google.com/p/gpuwire/.
The GPU API used was NVidia's CUDA.



I'm currently making small corrections to the thesis text as well as trying to clean up the code.
It's interesting to notice that the GUI used is Qt. I'll probably post an entry explaining how to glue CUDA and Qt.
If you're interested, please leave me a comment.

WebCam + OpenGL + OpenCV head tracking = Immersive 3d environment

UPDATE! This sample has evolved to a project sponsored by Google. Please check http://code.google.com/p/ehci

This is my first attempt to create a 3d immersive environment using a simple webcam to track head's position.



This video shows my approach to combine OpenCV head tracking with a 3d OpenGL environment so that the user will see objects from their head's point of view. Since head tracking is 2d, no depth information is obtained. Besides that, it takes the face recognition algorithm around 200ms, which yields low fps (about 5). I'm currently trying to improve that :)

*******
Well, thanks to Walter Piechulla I could decrease detection time to around 20ms using the flag CV_HAAR_FIND_BIGGEST_OBJECT - which makes opencv detect only one face - in a parameter to the function cvHaarDetectObjects, that looks like this now:

cvHaarDetectObjects( small_img, cascade, storage, 1.1, 2, CV_HAAR_FIND_BIGGEST_OBJECT,
cvSize(30, 30) );

(by the way, I had to update OpenCV from CVS)
As soon as I get some time, I can make another video showing speedup :D
Thanks Piechulla

********
The new version is available here: http://danielbaggio.blogspot.com/2008/03/enhanced-version-of-head-tracking-and.html

Avoiding ACM Web Account to download papers

In case you are a researcher and you really want that paper but ACM web account is blocking you from seeing it, you might try the following:

• Access the paper from your university - some universities have granted access because of their IP, which makes it very easy for researchers to download the papers
• Get yourself an ACM Web Account
• In case you are a researcher and you simply are not at the university when trying to see the paper, an useful workaround can be looking for the author's homepage or the congress that published the article

For instance, I was looking for the article "Early Experience with Scientific Programs on the Cray MTA-2" which leads me to this page, which doesn't allow me to get the file. Then, googling for the first author and a related word "Wendell Anderson cray mta 2" leads me to the Congress page, from where I can get it. It generally works :)
Besides that, ACM Web Account seems to be a great service. I would really recommend getting one account :)
Hope this post can help you!


************** this text can be useful for google indexing *******************
Full-Text is a controlled feature.
To access this feature:

• Please login with your ACM Web Account.
• Please review the requirements below.

From OpenSocial's sandbox

I've been playing out with OpenSocial's API from within Orkut's sandbox for a while. This API intends to integrate social sites like Orkut, MySpace, Hi5 and others through a javascript API.
In order to test, I've signed up for a sandbox account and started reading the documentation api .
After creating the Hello Worlds gadgets - by the way, in order to use Orkut's sandbox, Google gadgets need to be created - I've came across the output on the right.
I'm still wondering why so few friends are retrieved... maybe it's a sandbox limitation... I don't know... I've just posted this to find it out.
****
Ooops... they've just answered that... only friends who are also in the sandbox are listed... that's why it only retrieves a couple friends :)

Small ruby script to remotely run a program

In case one needs to remotely run a program from some ruby application, an easy way to do that is using Net-ssh. It remotely runs commands through ssh.
After installing the gem through a simple

gem install net-ssh

the following script can be run in order to call the 'ls' application, used to list directory's contents:

require 'rubygems'
require 'net/ssh'

session = Net::SSH.start('your-host','your-user', 'your-password')

session.process.open("ls") do | ls |
ls.on_stdout do | p, data |
puts "#{data}"
end
end

What you should know before diving into GPGPU study

Before choosing your Graduate/Master/PhD thesis subject as something in GPGPU, you should be aware that:

1- A new technology, from NVidia, called CUDA, has been created and you don't need to use Graphics API to make general purpose programming using GPUs. Besides being outdated, confusing and not productive, graphic API code lacks several features available from CUDA, specially the ability to make scatter. Unfortunatelly programs written for CUDA are only going to run in GeForce 8000 series (and later), QuadroFX 5600, 4600 and Tesla C870, D870, S870. Other frameworks as Brook and RapidMind should be checked as well.

2- You need to understand that a GPU, as you'll use it, is a set of SIMD (single instruction multiple data) processors grouped in multiprocessors. It means THEY WILL ALWAYS EXECUTE THE SAME INSTRUCTION during a clock cycle, but in different data. So, you should not think of them as being able to execute threads in a fashion like a dual-core Intel would, or even the Cell processor. Make sure your algorithm can be expressed in such a way.

3- Memory transfers can become potential bottlenecks for your application. Remember that, when programming GPU code, the data you access needs to be located at the global memory space in GPU. In order to load something there, you'll firstly need to download data to the GPU. As a rule of thumb, it takes a GeForce 8600M GT around 6ms to download a 1024x1024 pixels image (around 4MB). And YES, it might be the single detail that will prevent your application from running faster on GPUs.

4- There's some initialization time for the GPU (which is around 100ms). If you need to run only one iteration of your algorithm, and it is supposed to take less than 100 ms... CPU will probably be the best choice.

5- There's very small shared memory for GPUs. CUDA architecture makes available is 16KB per multiprocessor. It means, that if you are running 256 threads in a single multiprocessor (which is what you should be generally doing) you'd have 64 Bytes per thread... which are about 16 integers... if you can solve your problem with such a small amount of memory, your application will be extremely fast, because shared memory access is almost as fast as register access. Else, you'll need to write to global memory... it takes around 300 cycles to retrieve a value from global memory - which has around 256MB depending on your GPU. If you cannot hide this latency processing the retrieved data while other requests are made, your application will really lag in this point. A good strategy is to allow other threads to process the data while others are waiting for memory requests. An application heavily focused on memory requests will suffer dramatically because of this.

6- Only single precision floating point registers are available. It means around 7 digits of mantissa. Of course, future GPU generations will include double precision values.

7- Linear algebra problems will most probably run very fast on GPUs.

8- Large arrays, such as N-Body simulations are also a good fit for GPUs.

Some benchmarks on CPU, Cell and GPUs are included in this paper. It should be noticed, though, that the article hasn't been accepted for the congress, mainly because this application really does not require much processing from Cell or GPUs, as it spends most of the time on memory transfers. For more info, a great article about the subject and a great site www.gpgpu.org.

*Source: NVidia CUDA Programming Guide and personal experience with GPGPU

latex error .sty not found

Hi guys, if you've come into this error:
"latex error your-new-file.sty not found"
while building your .tex files, and you are sure you've just put them in some sort of directory like:
/usr/share/texmf/tex/latex/your-new-directory
you should run the command
$ texhash
and then, latex should be ready to find your style files
:)

Imobiliária São José dos Campos

Aluguei um excelente imóvel na milimoveis www.milimoveis.com.br milimoveis.blogspot.com

Good friends @ Northwestern University - Robert H. Lurie Comprehensive Cancer Center

Well, just for the records, I'd like to leave this picture here, with all the great colleagues I've met in Northwestern University. Special thanks to Simon, Dong, Rhett and Pan Du for the help along the project :)

Summer of Code final project

GeneQuad, an automatic way to describe genes based on their GeneRif (Gene Reference into function) is available through a rails web application, in the following address:
GeneQuad
The other project, GenDex is also available.
Both of them are intended to be useful in cancer research field as well as bio-informatics.
Guess what's the most researched gene in 2007?

Rising Stars

I've made some research finding out the genes that had more importance according to published articles in the last years. The findings are reported in these pages http://code.google.com/p/genequad/wiki/RisingStars and http://code.google.com/p/genequad/wiki/DetailsRisingStars

On the other hand, the ones that had less importance are also reported at http://code.google.com/p/genequad/wiki/ShowStoppers

Importing pubMed database with a ruby script using bioruby

If you desire to import pubMed database, a useful library for ruby may be used: bioruby. This post http://code.google.com/p/genequad/wiki/ImportingPubMed shows how to do that.

Zipf's Law in geneRIF

This post http://code.google.com/p/genequad/wiki/ZipfsLaw shows a study about Zipf's law in geneRIF.

Importing geneRIF database to sqlite3

Ok, if someone wants to import the txt file from NCBI to sqlite3, http://code.google.com/p/genequad/wiki/ImportingGenerif tells you how to do that. A procedure just like this may be used to import to other databases as well. Congrats NCBI for geneRIF!

Some studies about Zipf's law and genes

Well, actually, I only needed somewhere to upload this picture :)
If you want to check my reports about the genequad project on Summer of Code, you'll get a good clue of what it's about :)

Trick to change octave's plot background

Well, I was trying to change octave's plot background, but I found no easy way, so I changed the way gnuplot is called (gnuplot is the program octave calls to plot the graphics). Simply go to /usr/bin, log as root, and backup gnuplot as

cp gnuplot gnuplot.bak

Rename it to gnuplot1:

mv gnuplot gnuplot1

Now create the script gnuplot by typing

emacs gnuplot

Create this script:

#!/bin/bash
/usr/bin/gnuplot1 -background white

Add execution permission through chmod +x gnuplot

It will make a standard call to gnuplot change to gnuplot -background white, which will change your plots background in octave.
If you have a better way to do this, please leave me a comment
Have fun :)

IVUS Plugin available

In order to make my graduation thesis fully repeatable, I'm making the IVUS plugin available. This ImageJ plugin adds the functionality to slice an image frame sequence in 4 directions, so that longitudinal segmentation may be applied. It is used for Coronary Intravascular ultrasound images.
The download is available here http://sourceforge.net/project/showfiles.php?group_id=171446&package_id=195893&release_id=475672 under the ivussnakes package in IVUS Plugin release.
Enjoy it!

IBM Cell Workshop with Peter Hofstee

So, these days have been great!!! I was able to present a purpose from ITA to Peter Hofstee!!! I just can't believe such a genius was here in Brazil and we had the opportunity to interact with him in two days of workshops.
Besides him, Bob Szabo was there and I was able to show him www.cellperformance.com. It was great!
I'll be able to post some of what happened these days in the following posts! Keep tuned!

My Graduation Thesis is Online

Well, I've finally made my graduation thesis available online. I hope it may help someone :)
The link is here: http://downloads.sourceforge.net/ivussnakes/TG.pdf
The abstract is the following:

This graduation thesis deals with intravascular ultrasound (IVUS) image segmentation through LiveWire and Snakes algorithms. As the type of ultrasound studied is the one from coronaries, there’s an initial detailing of the medical environment in which the problem is inserted.
The introduction is followed by an explanation of how LiveWire technique works, which is by calculating the shortest cost path between two pixels of the image. There’s a discussion about the implementation, using binary trees and a heap data structure. This is followed by the evaluation of which costs should be adopted, which were: modulus and direction of the gradient, the laplacian and a non-linear function.
Then, Snakes algorithm is described, which simulates a set of vertices submitted to one internal field of energy and another external one. There’s also a discussion about the number of iterations to be adopted as well as the values used as parameters.
It is explained, so, how simulated images, used for evaluation, were generated. Besides, a model for Speckle noise is described as well as the filter to which the images were submitted.
There’s a brief discussion on how the algorithms were implemented, with a special focus on ImageJ platform and it’s extensibility features.
The results are, then, described, with each of the simulated image types: noiseless, with Speckle noise and filtered. The main results found are that LiveWire has better segmentation quality and is slightly affected by noise, while Snakes is faster and is practically not affected by wrong initial points retrieved from longitudinal segmentation.
There’s, then, the conclusion, describing what was found, focusing the high velocity of Snakes technique, followed by good results. Although better results were found for LiveWire, the processing time was about 500 times higher, what might indicate this method is not desirable. Due to the good results found with simulated images, the developed platform promises to work efficiently in a medical environment, after submitted to validation by specialists.

I've got more news! Stay tunned!