First off: Well done for wanting to contribute!
I'd say the first step is to spend some significant time in reading and understanding the code of the project you want to contribute on. Not only on how it implements a certain algorithm, but also on how the project uses templates, inheritence and coding conventions (as these may change from project to project).
Since you already know the basics, this will teach you the application of these techniques in the real world, and how they are used in this project.
If you want some more in-depth background reading which may guide you in understanding the techniques used in the project, i'd suggest "Design Patterns" by Gamma, Helm, Johnson, and Vlissides http://en.wikipedia.org/wiki/Design_pattern_(computer_science)
Sorry for my late reply.
It is a university in the Netherlands.
I tried to double-check, but i can't find the exact rules anymore.
I do remember it specifically, because i wanted to put code written during my master's thesis online but had to ask permission from the university.
I guess you can compare it to doing graduation work within a company, the company owns the work the student does. The same i guess holds for the university.
Yes, moving outside the pure math may be a good idea.
I've learned the most math when i finally understood what it could be used for. A practical side helps.
For me, "Multiple View Geometry", by Hartley and Zisserman falls in this category. Well written, and uses nice mathematical tools to a very practical problem: 3d reconstruction in computer vision.
So, any fans of the SF/Fantasy genre out there who can say why Otherland?
Why not Steven Erikson's Malazan Empire? Robert Jordan's Wheel of (too much) time, Tolkien, or Terry Pratchett for that matter?
What makes Otherland more suitable than the others?
I find it interesting which ones of the object-recognition and scene categorization algorithms make it to Slashdot. Why does this one make it? This is a very hot research topic at the moment. to name a couple of groups:
oh, and people should not stare themselves blind on the claimed results. Research papers *always* have to present good results, or else you do not get published. Furthermore, these images are of a very high quality, make by professional photographers. Many algorithms perform very well on these ('corel'-like) sets, while utterly failing if applied on real-world data: http://www-nlpir.nist.gov/projects/trecvid/
I think there is a difference between a 'normal' startup and an internet-startup.
This difference is between the non-tangibleness of the internet (and software in general) and a production company.
You can start developing a website in your spare time, and the end result doesn't take up 5 factory floor.
I would be very interested in your research, can you post some pointers to modeling feedback?
The caltech datasets are in my opinion artificial, since they rotate all images in the same direction. For example, a moterbike always faces to the right, and the 'trilobite' is even rotated out of the plane (leaving a white background) so you only need to estimate the right angle of rotation. for example, see: http://www.vision.caltech.edu/Image_Datasets/Calte ch101/averages100objects.jpg you would never get a consistent blurred image if you would allow unconstrained views of an object.
This paper's claim to recognize scenes like the brain does, is overdrawn. As far as i can tell from their paper (it is a journal version of their cvpr paper) only their low-level Gabor features are similar to what the brain does. The rest of the paper uses the currently popular bag-of-features model, which is a model that discards all spatial information between image features, which i don't think the brain does. Furthermore, for classification algorithms they consider a Support Vector Machine and Boosting. Both of these classifiers are certainly not comparable to what the brain does. Why not use a neural network if they aim is to mimic the brain? Furhtermore, they only conside feed-forward information, where research shows that there is at least as much information going back as there is going forward.
Don't get me wrong, it is still a nice paper, with good results. (however, all Caltech datasets are highly artificial, with objects artificially rotated in 1 direction) So, nice paper, but to compare it with the workings of the human brain is too much.
I'm sorry, but that pdf link you posted, is not a scientific article, it seems like a requirements specification made by business students. The science in your 'prior art' is mostly segmentation, and has very limited validation.
Wait a minute, not all Flash is bad, right? ok, if it's annoying it's probably Flash or animated Gif (thats why firefox has Adblock) However, the use of Flash doesn't *have* to be annoying? does it?
would people object against a well designed Flash site? Flash can do so much more than plain HTML (olthough usually used for bad things, though)
I can't find the answer in the free sample of this book (to make it a little more on-topic) it does mention useless splash screens, and using text besides Flash for search engine support.
check the TRECVID Video retrieval Benchmark.
(not a compitition) but a directed research effort to make searching video possible.
It was a spin-off from the Text REtrieval Convference The driving force of text retrieval.
TrecVID has a retrieval task, ie, "find shots of people walking up stairs", "find shots of Bill Clinton with an american flag in the background"
And a semantic concept-detection task, ie, train on a test set, and try to detect sport videos, bicycles, buildings.
Video retrieval is still in its infancy. So, processing the text from the speech recognizer still provides the most information.
btw.. high correleation between image and text is a hard issue. Especially since computer vision is a very hard problem.
But! check the paper of the (um, ok, it's also my) University of Amsterdam, who "won" (not a compitition;) this year's TrecVID benchmark, and learn about the Semantic Value Chain
note that Google hasn't participated in any of the TRECVID benchmarks. IBM seems to be the only big (well performing) commercial participant.
I like OCaml because it Combines the power of functional programming, like (tail-)recursion, functions as an argument, with 'normal' programming language statements. It doesn't force the "functional programming way" on you, like Lisp does, So, you can still use the If then, and While statements if you find them more usefull then a recursion.
And, it is quite fast! both in development and in execution. however, i have yet to find a way to well commenting my code.
Slashdot Mirror
Smells a bit fishy to me: http://boingboing.net/2012/10/02/what-a-dead-fish-can-teach-you.html
First off: Well done for wanting to contribute!
I'd say the first step is to spend some significant time in reading and understanding the code of the project you want to contribute on. Not only on how it implements a certain algorithm, but also on how the project uses templates, inheritence and coding conventions (as these may change from project to project). Since you already know the basics, this will teach you the application of these techniques in the real world, and how they are used in this project.
If you want some more in-depth background reading which may guide you in understanding the techniques used in the project, i'd suggest "Design Patterns" by Gamma, Helm, Johnson, and Vlissides http://en.wikipedia.org/wiki/Design_pattern_(computer_science)
Sorry for my late reply.
It is a university in the Netherlands.
I tried to double-check, but i can't find the exact rules anymore.
I do remember it specifically, because i wanted to put code written during my master's thesis online but had to ask permission from the university.
I guess you can compare it to doing graduation work within a company, the company owns the work the student does. The same i guess holds for the university.
Data Center Overlords?
This is probably not the case. At my University they own the rights to all code that is written as part of the educational program, not the students.
I recommend the masters, a higher degree will make it easier to switch jobs later in your career
Yes, moving outside the pure math may be a good idea. I've learned the most math when i finally understood what it could be used for. A practical side helps. For me, "Multiple View Geometry", by Hartley and Zisserman falls in this category. Well written, and uses nice mathematical tools to a very practical problem: 3d reconstruction in computer vision.
So, any fans of the SF/Fantasy genre out there who can say why Otherland? Why not Steven Erikson's Malazan Empire? Robert Jordan's Wheel of (too much) time, Tolkien, or Terry Pratchett for that matter? What makes Otherland more suitable than the others?
yes, to program such an interface is of course very easy, just as simple as writing a program to check if a thread ever halts.
I find it interesting which ones of the object-recognition and scene categorization algorithms make it to Slashdot.
m
Why does this one make it?
This is a very hot research topic at the moment.
to name a couple of groups:
http://www.robots.ox.ac.uk/~vgg/
http://lear.inrialpes.fr/
http://www.vision.caltech.edu/
http://www.science.uva.nl/research/isla/
http://www.cdvp.dcu.ie/
http://www.informedia.cs.cmu.edu/
http://www.research.ibm.com/slam/
http://www.ee.columbia.edu/ln/dvmm/newResearch.ht
oh, and people should not stare themselves blind on the claimed results.
Research papers *always* have to present good results, or else you do not get published.
Furthermore, these images are of a very high quality, make by professional photographers.
Many algorithms perform very well on these ('corel'-like) sets, while utterly failing if applied on real-world data:
http://www-nlpir.nist.gov/projects/trecvid/
I think there is a difference between a 'normal' startup and an internet-startup. This difference is between the non-tangibleness of the internet (and software in general) and a production company. You can start developing a website in your spare time, and the end result doesn't take up 5 factory floor.
I would be very interested in your research, can you post some pointers to modeling feedback?
e ch101/averages100objects.jpg
b ases.html#VOC2006
The caltech datasets are in my opinion artificial, since they rotate all images in the same direction.
For example, a moterbike always faces to the right, and the 'trilobite' is even rotated out of the plane (leaving a white background) so you only need to estimate the right angle of rotation.
for example, see:
http://www.vision.caltech.edu/Image_Datasets/Calt
you would never get a consistent blurred image if you would allow unconstrained views of an object.
Better datasets in my opinion are the VOC challenge:
http://www.pascal-network.org/challenges/VOC/data
and the digital video benchmark Trecvid (where we work on)
http://www-nlpir.nist.gov/projects/trecvid/
which is not only true real-world data, it consists of hundereds of hours of video, instead of a few of thousand images.
This paper's claim to recognize scenes like the brain does, is overdrawn.
As far as i can tell from their paper (it is a journal version of their cvpr paper) only their low-level Gabor features are similar to what the brain does.
The rest of the paper uses the currently popular bag-of-features model, which is a model that discards all spatial information between image features, which i don't think the brain does. Furthermore, for classification algorithms they consider a Support Vector Machine and Boosting. Both of these classifiers are certainly not comparable to what the brain does. Why not use a neural network if they aim is to mimic the brain?
Furhtermore, they only conside feed-forward information, where research shows that there is at least as much information going back as there is going forward.
Don't get me wrong, it is still a nice paper, with good results.
(however, all Caltech datasets are highly artificial, with objects artificially rotated in 1 direction)
So, nice paper, but to compare it with the workings of the human brain is too much.
I'm sorry, but that pdf link you posted, is not a scientific article, it seems like a requirements specification made by business students.
l
The science in your 'prior art' is mostly segmentation, and has very limited validation.
If you say Kobus Barnard's work has prior art, that is true, because his work is very related to the ALIP system.
see http://kobus.ca/research/talks/INRIA-02/index.htm
Wait a minute, not all Flash is bad, right?
ok, if it's annoying it's probably Flash or animated Gif (thats why firefox has Adblock)
However, the use of Flash doesn't *have* to be annoying?
does it?
would people object against a well designed Flash site?
Flash can do so much more than plain HTML (olthough usually used for bad things, though)
I can't find the answer in the free sample of this book (to make it a little more on-topic)
it does mention useless splash screens, and using text besides Flash for search engine support.
It was a spin-off from the Text REtrieval Convference The driving force of text retrieval.
TrecVID has a retrieval task, ie, "find shots of people walking up stairs", "find shots of Bill Clinton with an american flag in the background" And a semantic concept-detection task, ie, train on a test set, and try to detect sport videos, bicycles, buildings.
Video retrieval is still in its infancy. So, processing the text from the speech recognizer still provides the most information.
btw.. high correleation between image and text is a hard issue. Especially since computer vision is a very hard problem.
But! check the paper of the (um, ok, it's also my) University of Amsterdam, who "won" (not a compitition ;) this year's TrecVID benchmark, and learn about the Semantic Value Chain
note that Google hasn't participated in any of the TRECVID benchmarks. IBM seems to be the only big (well performing) commercial participant.
but there are!
I learned the language online, (just as i learned Java from their very good tutorial)
they gave 2 good ones at the top.
the Richard Jones'
Ocaml Tutorial for people who know how to program 'normally'
There is also an update to Jason Hickey's book
I like OCaml because it Combines the power of functional programming, like (tail-)recursion, functions as an argument, with 'normal' programming language statements.
It doesn't force the "functional programming way" on you, like Lisp does, So, you can still use the If then, and While statements if you find them more usefull then a recursion.
And, it is quite fast!
both in development and in execution.
however, i have yet to find a way to well commenting my code.