that wasn't my problem. my issue was trying to figure out what the hell it means to assign a/question/ a truth value. i can understand what it means when you're operating on statements, but the semantics of an 80% correct answer to a 5% true question kinda boggles me.
as a random fun fact, there are at least 13 references to semen in their list of facts and several random references to sex acts. i suspect that when gac grows up, it's going to turn into the average irc luser...
"`For this to work, it has to be on hundreds of millions of desktops so there's an improved consumer experience in advertising, search and content,' Eagle said."
what better way to get it onto the desktops via MSN? you get verizon broadband customers for free, plus those foolish enough to use microsoft's msn client. the implications...
so revolutionary, in fact, that winamp had been doing it for years before apple hit upon the idea. really, why sue apple when these guys could take on some far meatier offenders (aol, who owns nullsoft, the guys who made winamp, and m$). oh yeah... i'm thinking rationally again.
i have mixed feelings on opencv -- it feels really halfassed a lot of the time (e.g: edge detection only works in grayscale). i'm also really peeved by the documentation (or, in most cases, lack thereof).
that said, when it does what you want, it works great. unfortunately, they insisted on using C rather than C++, thus placing really arbitrary (and, imho, dumb) limits on the kinds of images you can process. there's no reason for 99% of the algorithms to care about what kinds of pixel types they're working on. logically, some operations only make sense on certain types of images, but to spend my life converting between formats is kinda dumb -- this is one case where using templates would have made *so* much more sense, oh well.
i don't understand why this entire discussion is framed in the dialog of os holy wars. i use linux at work because it's appropriate for what i do (robotics, machine vision). i use a mac at home because it doesn't suck as a desktop os. in fact, after many years of linux / solaris / (*shudder*) tru64 / several generations of windows, my experiences with os x have been by far the most pleasant. things just work (tm). does that mean i will forsake linux? not likely -- linux is a/tool/ which is good when used appropriately, just like os x (and, i suppose, even windows).
first off, there's intel's open computer vision library (check http://sourceforge.net/projects/opencvlibrary/) . you'll find a large chunk of the building blocks you'd want to build your own algorithms there (edge detection, line extraction, math primitives, etc.).
secondly: yes, there are ways to improve the performance of these algorithms. in general, the higher the resolution, the better (assuming you have the time to do the processing). if i recall correctly, most medical images use 16 bit gray scale at really high resolutions. in general, higher resolution at greater bit depths will help quite a bit -- it's all about signal to noise ratios.
it's a neat process. before i read the article i'd pictured an inkjet-esque approach. probably a good thing they didn't go that way --- can you imagine how much consumables would cost? to say nothing of issues related to poor quality drivers...
they intend to build a secure national id system out of technologies which have proven themselves to be insecure at each turn?
god forbid there ever be something like code red or equivalent that hits this system, because the resulting sound will be that of 280 odd million people being simultaneously sodomized by very large cacti.
it's free, available here: http://bugzilla.org/ it's not exactly designed for your specific task (it's primarily aimed at quality control / bug tracking) but shouldn't be hard to adopt for your purposes.
this is exactly what i was going to suggest, but you beat me to it.
html email is really a waste of bandwidth. do what acm does: email out a quick summary with links, and i'll go peruse as my fancy takes me. no need to waste all that extra bandwidth with formatting (or cpu time with compression). almost all html email that gets sent to me is immediately whisked away to my trash can (there are a few people who i actually want to hear from that still insist on sending html email, there are exceptions to my filtering rules for them).
i also just finished a master's degree in comp. sci. (focussing in computer vision). i then landed what is essentially my dream gig at r&d driven firm. i spend quite a bit of time seeing what citeseer and acm's digital library have to say on matters of interest (i'm not an ieee member, although i should eventually look into it). typically, i either end up saving a huge amount of time eliminating approaches or i end up with cool ideas on how to combine approaches, which has thus far been really useful. your mileage will vary, of course.
... when it goes backwards. at least i think it does (this is a very vague recollection). something to do with split light beams and what not, or it may be in "a brief history of time."
pi itself is not infinite, the number of digits necessary to represent it, however, are.
the fact that pi is a very special number (not only irrational but transcendental, and thus not the root of any integer polynomial) makes the randomness assumption very tempting.
... but not for cheap as far as i know. we have played with several imus and a couple accelerometers. all the reasonably cheap ones have enormous drift rates, and the rest all seem to be $1500+.
that said, once you have the sensor, you can talk to it pretty easily using a standard serial interface.
...amongst the many fun toys at work is a truck with about 20U rack space in the trunk, power generators, etc. --- the perfect venue for the world's fastest lan party...
the claim that a city like san francisco is going to be totally unable to handle the logistics of wifi is, well, ridiculous. cities have to juggle a lot more than phone networks: they have to handle the logistics of roads, libraries, health services, schools, etc. --- a task which in my totally uneducated opinion appears to be substantially more complicated than running a wifi network.
the rest of the article seems to serve only as proof that seidenberg and the industry he serves is full of proud egomaniacs.
i suppose i shouldn't be too surprised that a slashdot editor didn't bother to read the article they're posting, but i'd like to point out that in this case the problem was *not* a university being careless about data. the problem is that a student, by abusing her access to confidential data, was able to gain access to the same shared secrets that were used to authenticate network users. to the university's credit, they had an audit system in place which caught the problem.
spread spectrum is pretty cool, but to be/really/ effective, i'd suggest throwing in some steganography. for example, instead of choosing a single frequency at a time, you pick three and simultanenously transmit on all three channels. you choose one of those channels as the 'true' channel, and (using a stochastic process of some sort, i'd imagine), manufacture plausible signals on the other two channels that are totally ignored receiving side. joe average will pick up that you're using spread spectrum, but even if they do they've still got a nasty problem of figuring out which of each set of three signals is really the right one to use, especially since all three signals will be essentially undistinguishable in terms of information content.
the other option is to use a more interesting where the signal is encoded but as bursts in a kind of morse-code-y way. e.g: transmit several symbols on the channel, but only choose one of them to be significant; designate a single occurance of that symbol as a dot and two adjacent occurances as a dash, require that all dots and dashes be separated by some other symbol. lots of things you can do once you bring steg into play.
Slashdot Mirror
grep | wc -l is my friend.
that wasn't my problem. my issue was trying to figure out what the hell it means to assign a /question/ a truth value. i can understand what it means when you're operating on statements, but the semantics of an 80% correct answer to a 5% true question kinda boggles me.
as a random fun fact, there are at least 13 references to semen in their list of facts and several random references to sex acts. i suspect that when gac grows up, it's going to turn into the average irc luser...
"`For this to work, it has to be on hundreds of millions of desktops so there's an improved consumer experience in advertising, search and content,' Eagle said."
what better way to get it onto the desktops via MSN? you get verizon broadband customers for free, plus those foolish enough to use microsoft's msn client. the implications...
so *that's* why they're called lusers...
i already get spammed by m$ and its various subsidiaries as it is. i really don't need them to get any better at it than they already are.
so revolutionary, in fact, that winamp had been doing it for years before apple hit upon the idea. really, why sue apple when these guys could take on some far meatier offenders (aol, who owns nullsoft, the guys who made winamp, and m$). oh yeah... i'm thinking rationally again.
i have mixed feelings on opencv -- it feels really halfassed a lot of the time (e.g: edge detection only works in grayscale). i'm also really peeved by the documentation (or, in most cases, lack thereof).
that said, when it does what you want, it works great. unfortunately, they insisted on using C rather than C++, thus placing really arbitrary (and, imho, dumb) limits on the kinds of images you can process. there's no reason for 99% of the algorithms to care about what kinds of pixel types they're working on. logically, some operations only make sense on certain types of images, but to spend my life converting between formats is kinda dumb -- this is one case where using templates would have made *so* much more sense, oh well.
i don't understand why this entire discussion is framed in the dialog of os holy wars. i use linux at work because it's appropriate for what i do (robotics, machine vision). i use a mac at home because it doesn't suck as a desktop os. in fact, after many years of linux / solaris / (*shudder*) tru64 / several generations of windows, my experiences with os x have been by far the most pleasant. things just work (tm). does that mean i will forsake linux? not likely -- linux is a /tool/ which is good when used appropriately, just like os x (and, i suppose, even windows).
first off, there's intel's open computer vision library (check http://sourceforge.net/projects/opencvlibrary/) . you'll find a large chunk of the building blocks you'd want to build your own algorithms there (edge detection, line extraction, math primitives, etc.).
secondly: yes, there are ways to improve the performance of these algorithms. in general, the higher the resolution, the better (assuming you have the time to do the processing). if i recall correctly, most medical images use 16 bit gray scale at really high resolutions. in general, higher resolution at greater bit depths will help quite a bit -- it's all about signal to noise ratios.
hope this helps
it's a neat process. before i read the article i'd pictured an inkjet-esque approach. probably a good thing they didn't go that way --- can you imagine how much consumables would cost? to say nothing of issues related to poor quality drivers...
as another of those 280 million, i hope i'm speaking metaphorically too. that said, you won't catch me near the desert any time soon ;-)
they intend to build a secure national id system out of technologies which have proven themselves to be insecure at each turn?
god forbid there ever be something like code red or equivalent that hits this system, because the resulting sound will be that of 280 odd million people being simultaneously sodomized by very large cacti.
it's free, available here: http://bugzilla.org/ it's not exactly designed for your specific task (it's primarily aimed at quality control / bug tracking) but shouldn't be hard to adopt for your purposes.
this is exactly what i was going to suggest, but you beat me to it.
html email is really a waste of bandwidth. do what acm does: email out a quick summary with links, and i'll go peruse as my fancy takes me. no need to waste all that extra bandwidth with formatting (or cpu time with compression). almost all html email that gets sent to me is immediately whisked away to my trash can (there are a few people who i actually want to hear from that still insist on sending html email, there are exceptions to my filtering rules for them).
i also just finished a master's degree in comp. sci. (focussing in computer vision). i then landed what is essentially my dream gig at r&d driven firm. i spend quite a bit of time seeing what citeseer and acm's digital library have to say on matters of interest (i'm not an ieee member, although i should eventually look into it). typically, i either end up saving a huge amount of time eliminating approaches or i end up with cool ideas on how to combine approaches, which has thus far been really useful. your mileage will vary, of course.
... when it goes backwards. at least i think it does (this is a very vague recollection). something to do with split light beams and what not, or it may be in "a brief history of time."
forgive the nitpicking:
pi itself is not infinite, the number of digits necessary to represent it, however, are.
the fact that pi is a very special number (not only irrational but transcendental, and thus not the root of any integer polynomial) makes the randomness assumption very tempting.
who makes several tiger branded motherboards...
... but not for cheap as far as i know. we have played with several imus and a couple accelerometers. all the reasonably cheap ones have enormous drift rates, and the rest all seem to be $1500+.
that said, once you have the sensor, you can talk to it pretty easily using a standard serial interface.
if only i knew what the greek for irrational fear of fear is ;-)
glad to see i'm not the only person who was bothered by that statement. guess it'd be called something like a triskadecahedron.
...amongst the many fun toys at work is a truck with about 20U rack space in the trunk, power generators, etc. --- the perfect venue for the world's fastest lan party...
the claim that a city like san francisco is going to be totally unable to handle the logistics of wifi is, well, ridiculous. cities have to juggle a lot more than phone networks: they have to handle the logistics of roads, libraries, health services, schools, etc. --- a task which in my totally uneducated opinion appears to be substantially more complicated than running a wifi network.
the rest of the article seems to serve only as proof that seidenberg and the industry he serves is full of proud egomaniacs.
i suppose i shouldn't be too surprised that a slashdot editor didn't bother to read the article they're posting, but i'd like to point out that in this case the problem was *not* a university being careless about data. the problem is that a student, by abusing her access to confidential data, was able to gain access to the same shared secrets that were used to authenticate network users. to the university's credit, they had an audit system in place which caught the problem.
spread spectrum is pretty cool, but to be /really/ effective, i'd suggest throwing in some steganography. for example, instead of choosing a single frequency at a time, you pick three and simultanenously transmit on all three channels. you choose one of those channels as the 'true' channel, and (using a stochastic process of some sort, i'd imagine), manufacture plausible signals on the other two channels that are totally ignored receiving side. joe average will pick up that you're using spread spectrum, but even if they do they've still got a nasty problem of figuring out which of each set of three signals is really the right one to use, especially since all three signals will be essentially undistinguishable in terms of information content.
the other option is to use a more interesting where the signal is encoded but as bursts in a kind of morse-code-y way. e.g: transmit several symbols on the channel, but only choose one of them to be significant; designate a single occurance of that symbol as a dot and two adjacent occurances as a dash, require that all dots and dashes be separated by some other symbol. lots of things you can do once you bring steg into play.