True but he was only using a single node, and it cost him $2 or something. He could easily use a few hundred nodes and get that "SIX MONTHS" down to a few hours.
The important technology in the Kinect is supplied by PrimeSense.
I know.
Their web site describes the operation of the depth-sensing system, including the Structured Light projection used
No it doesn't. However, thanks for the "Structured Light" keyword - wikipedia has an article about it with some links. It seems that it is just triangulation of the dots. Pretty amazing that it works as well as it does!
It's not just a webcam. It's some kind of laser IR projector (NOT LEDs), and a monochrome IR webcam. The laser emits lots of points (see videos on youtube), which are somehow 'coded' according to primesense, and then the webcam somehow turns this into depth information.
I still haven't seen an explanation of that step though, and I can't work it out. Surely it's not just triangulation? It obviously isn't time-of-flight as some people have suggested.
By the way, I just had a great idea for how to get two Kinects to point at the same area in a stereo configuration - use active 3D glasses!
Utter bullshit. That's just the BOM for the 'major' chips. It doesn't include PCBs, small components like capacitors and voltage regulators, the housing, lenses, cables, connectors, tilt servo. Nor does it include the cost of assembly, transport and packaging.
I suppose you think nice restaurants are a rip-off because the price of eggs and flour is so low.
Actually, only one is really simple. I think my 1.5 year old son would love to see a big rabbit on TV that would simply replicate his every move. Not much of a game, but a cool tech demo, as well as something for my son to enjoy.
That's not remotely simple. All the pose-tracking stuff is done in the xbox. It's complicated computer vision stuff, and it will be *extremely* hard to get it to work as well as Microsoft has.
The other (much more ambitious) idea, is to mix it with an HTML 5 demo I already was considering. I'd need some way to turn Kinect events into mouse events, I guess. Something that a browser can handle, in any case, so I think that means mouse events. Something multitouchy would be nice, but I don't think browsers support that, do they?
That would be much much easier. I guess Google TV's browser might support multitouch, since it is just Android.
I doubt there's anything in the driver that depends on Windows. The fact that this guy managed to do this in less than a week indicates that microsoft haven't done anything to make Kinect hard to use. I expect the video streams are exposed as UVC devices.
Indeed, for instance firesheep could just use a different internet connection (e.g. 3G). Some websites check the source IP of the cookie, but most probably don't.
Embarrassing, sure. But it's just their website, and doesn't justify spending £500m on fighting "cyber-terrorism". By the way does anyone know what the £500m will actually be spent on? It *should* be spent on researching secure systems like BitC, SELinux, stack protection and so on. I bet it isn't.
"Steal: To take (the property of another) without right or permission."
It doesn't have to be physical property. One can steal ideas, research and designs. In fact designs are a very good example. If China steals the designs for a fighter jet from the US, the US still has the designs; they haven't lost anything. But it's still called stealing.
I'm not saying that piracy is the same as stealing an actual object, but piracy is still stealing. *You* may want to change the meaning of the word, but the rest of the world would call piracy stealing. I'll admit 'theft' has a slightly more specific meaning, but only slightly and it is perfectly acceptable to talk about the theft of fighter jet designs by Chinese spies.
it will be a bit pointless if all you get is a few mono images. You could knock something up with a few webcams now that does it.
No that's exactly what they're trying to get, and no you can't do it with a few webcams. Kinect uses new hardware that isn't available anywhere else (at least not for anything like Kinect's price). Look it up.
That's completely difference. Kinect produces an actual depth man (like a z-buffer) of what it sees using infra-red time-of-flight measurements (apparently). It's *not* stereoscopic, and I doubt the actual kinect hardware even does any people tracking (that is probably in the xbox).
The only thing they changed was the kernel. Performance differences can only be due to changes in the kernel. In what way is that not a kernel benchmark?
It's much much cheaper to shoot in 3D than to do it in post-processing. I honestly don't know why he cares - you can always shoot the film in 3D and show it in 2D if you like.
Labview is 'visual programming' done in about the worst way possible. Even the UI is badly done.
I think something more like scratch would work, where instead of using ASCII sequences to represent tokens and blocks, graphical objects can be snapped together like a jigsaw. The code still looks something like a normal program, instead of a tangled mess of multi-layered balls of string like Labview. And there's no possibility for syntax errors, type mismatches, etc. The compiler would be much faster and simpler because it already has the AST loaded.
Of course it has downsides. Typing is fast, and easy to share online. It works well with VCSs. Still, I'd like to see a serious programming language done in that style (one could easily do C for example).
All the "if it was real" link says is that Sony doesn't comment on speculation and rumour. It really doesn't look like a render, and there's no real evidence it is fake. In fact I'd expect a fake to have fewer flaws (e.g. "Emergency calls only"/"No SIM card" surely wouldn't overlap in a fake).
If you'd need to do _lots_ of path manipulation and thus lots of escaping, then doing it in bash makes no sense.
Exactly my point. Pretty much all code involves path manipulation, and *robust* code will have to escape it properly. Since that is a hassle in bash, it's not the best choice for robust code.
And re scp, I meant:
scp some_file remote:"/tmp/Directory With Spaces"
I doubt most people would remember to run that path through sed unless they happened to notice it not working.
Shell scripting is fine for stuff that *only you* are going to use. It's just not robust enough for use in anything important, that more than one person might actually use. For example, handling paths with spaces is pretty damn hard - loads of scripts can't handle them.
The huge problem with their magnetic gears is that above a certain torque they will lose all grip. It's the same with stepper motors.
Cool as this is, what they are basically doing is embedding a set of small magnets into one piece of magnet. For example, the same "frictionless magnetic gear" could be easily created by sticking a lot of alternating N/S magnets on the outside of some wheels. It's pretty much exactly the same thing.
Not sure that link is relevant. It is about Zuckerberg hacking into *email* accounts, not Facebook accounts (which of course he doesn't need to hack into).
Also, not a good idea to back anything up with a link to the Daily Mail!:-)
It's extremely simple. Before Facebook, it was still considered weird to use your real name on a website. Most names on MySpace were like "johnnys123" rather than "John Smith". Obviously using real names is much more desirable, and one of the main reasons Facebook became popular. And the reason people were willing to use real names on Facebook was because you needed a.ac.uk or.edu email address to get an account, and only people from your uni could see your profile.
In a nutshell:
1. It was much more secure than the alternatives. 2. So people felt ok using real names and details, and allowing other people to see their profiles (because only people from the same uni could). 3. The use of real details made it much more friendly and useful.
There were other reasons too:
1. It didn't look like shit like MySpace. 2. Due to the.ac.uk/.edu requirement it wasn't filled with idiots. 3. Luck.
Slashdot Mirror
True but he was only using a single node, and it cost him $2 or something. He could easily use a few hundred nodes and get that "SIX MONTHS" down to a few hours.
Yeah but salt.
The important technology in the Kinect is supplied by PrimeSense.
I know.
Their web site describes the operation of the depth-sensing system, including the Structured Light projection used
No it doesn't. However, thanks for the "Structured Light" keyword - wikipedia has an article about it with some links. It seems that it is just triangulation of the dots. Pretty amazing that it works as well as it does!
It's not just a webcam. It's some kind of laser IR projector (NOT LEDs), and a monochrome IR webcam. The laser emits lots of points (see videos on youtube), which are somehow 'coded' according to primesense, and then the webcam somehow turns this into depth information.
I still haven't seen an explanation of that step though, and I can't work it out. Surely it's not just triangulation? It obviously isn't time-of-flight as some people have suggested.
By the way, I just had a great idea for how to get two Kinects to point at the same area in a stereo configuration - use active 3D glasses!
http://concentriclivers.com/misc/stereo_kinect.pdf
The parts have been shown to cost about $50
Utter bullshit. That's just the BOM for the 'major' chips. It doesn't include PCBs, small components like capacitors and voltage regulators, the housing, lenses, cables, connectors, tilt servo. Nor does it include the cost of assembly, transport and packaging.
I suppose you think nice restaurants are a rip-off because the price of eggs and flour is so low.
Actually, only one is really simple. I think my 1.5 year old son would love to see a big rabbit on TV that would simply replicate his every move. Not much of a game, but a cool tech demo, as well as something for my son to enjoy.
That's not remotely simple. All the pose-tracking stuff is done in the xbox. It's complicated computer vision stuff, and it will be *extremely* hard to get it to work as well as Microsoft has.
The other (much more ambitious) idea, is to mix it with an HTML 5 demo I already was considering. I'd need some way to turn Kinect events into mouse events, I guess. Something that a browser can handle, in any case, so I think that means mouse events. Something multitouchy would be nice, but I don't think browsers support that, do they?
That would be much much easier. I guess Google TV's browser might support multitouch, since it is just Android.
I doubt there's anything in the driver that depends on Windows. The fact that this guy managed to do this in less than a week indicates that microsoft haven't done anything to make Kinect hard to use. I expect the video streams are exposed as UVC devices.
Anyway, I want one.
Indeed, for instance firesheep could just use a different internet connection (e.g. 3G). Some websites check the source IP of the cookie, but most probably don't.
Embarrassing, sure. But it's just their website, and doesn't justify spending £500m on fighting "cyber-terrorism". By the way does anyone know what the £500m will actually be spent on? It *should* be spent on researching secure systems like BitC, SELinux, stack protection and so on. I bet it isn't.
"Steal: To take (the property of another) without right or permission."
It doesn't have to be physical property. One can steal ideas, research and designs. In fact designs are a very good example. If China steals the designs for a fighter jet from the US, the US still has the designs; they haven't lost anything. But it's still called stealing.
I'm not saying that piracy is the same as stealing an actual object, but piracy is still stealing. *You* may want to change the meaning of the word, but the rest of the world would call piracy stealing. I'll admit 'theft' has a slightly more specific meaning, but only slightly and it is perfectly acceptable to talk about the theft of fighter jet designs by Chinese spies.
it will be a bit pointless if all you get is a few mono images. You could knock something up with a few webcams now that does it.
No that's exactly what they're trying to get, and no you can't do it with a few webcams. Kinect uses new hardware that isn't available anywhere else (at least not for anything like Kinect's price). Look it up.
That's completely difference. Kinect produces an actual depth man (like a z-buffer) of what it sees using infra-red time-of-flight measurements (apparently). It's *not* stereoscopic, and I doubt the actual kinect hardware even does any people tracking (that is probably in the xbox).
The only thing they changed was the kernel. Performance differences can only be due to changes in the kernel. In what way is that not a kernel benchmark?
*Fewer*.
Jesus.
It's much much cheaper to shoot in 3D than to do it in post-processing. I honestly don't know why he cares - you can always shoot the film in 3D and show it in 2D if you like.
Labview is 'visual programming' done in about the worst way possible. Even the UI is badly done.
I think something more like scratch would work, where instead of using ASCII sequences to represent tokens and blocks, graphical objects can be snapped together like a jigsaw. The code still looks something like a normal program, instead of a tangled mess of multi-layered balls of string like Labview. And there's no possibility for syntax errors, type mismatches, etc. The compiler would be much faster and simpler because it already has the AST loaded.
Of course it has downsides. Typing is fast, and easy to share online. It works well with VCSs. Still, I'd like to see a serious programming language done in that style (one could easily do C for example).
How about you RTFA? It's obviously not specific to social networks.
All the "if it was real" link says is that Sony doesn't comment on speculation and rumour. It really doesn't look like a render, and there's no real evidence it is fake. In fact I'd expect a fake to have fewer flaws (e.g. "Emergency calls only"/"No SIM card" surely wouldn't overlap in a fake).
Either way, I want one!
If you'd need to do _lots_ of path manipulation and thus lots of escaping, then doing it in bash makes no sense.
Exactly my point. Pretty much all code involves path manipulation, and *robust* code will have to escape it properly. Since that is a hassle in bash, it's not the best choice for robust code.
And re scp, I meant:
scp some_file remote:"/tmp/Directory With Spaces"
I doubt most people would remember to run that path through sed unless they happened to notice it not working.
Wait, so your solution to "spaces are hard in bash" is: 1. It's not a problem with bash, 2. Use perl?
The most hideous example of spaces in bash is trying to send a file with spaces using scp. You have to manually escape them using sed.
Shell scripting is fine for stuff that *only you* are going to use. It's just not robust enough for use in anything important, that more than one person might actually use. For example, handling paths with spaces is pretty damn hard - loads of scripts can't handle them.
The huge problem with their magnetic gears is that above a certain torque they will lose all grip. It's the same with stepper motors.
Cool as this is, what they are basically doing is embedding a set of small magnets into one piece of magnet. For example, the same "frictionless magnetic gear" could be easily created by sticking a lot of alternating N/S magnets on the outside of some wheels. It's pretty much exactly the same thing.
Still cool though.
Bullshit. Unless you're calling the Chrome Adblock author a liar.
https://chrome.google.com/extensions/detail/gighmmpiobklfepjocnamgkkbiglidom
"New in version 2.0: Ads are actually blocked from downloading now, instead of just being removed after the fact!"
Not sure that link is relevant. It is about Zuckerberg hacking into *email* accounts, not Facebook accounts (which of course he doesn't need to hack into).
Also, not a good idea to back anything up with a link to the Daily Mail! :-)
It's extremely simple. Before Facebook, it was still considered weird to use your real name on a website. Most names on MySpace were like "johnnys123" rather than "John Smith". Obviously using real names is much more desirable, and one of the main reasons Facebook became popular. And the reason people were willing to use real names on Facebook was because you needed a .ac.uk or .edu email address to get an account, and only people from your uni could see your profile.
In a nutshell:
1. It was much more secure than the alternatives.
2. So people felt ok using real names and details, and allowing other people to see their profiles (because only people from the same uni could).
3. The use of real details made it much more friendly and useful.
There were other reasons too:
1. It didn't look like shit like MySpace. .ac.uk/.edu requirement it wasn't filled with idiots.
2. Due to the
3. Luck.