m****c
发帖数: 252 | 5 simple。。。搞计算机视觉的那帮人都得哭死，搞了几十年的computer vision既然是
个simple的engineering problem。
我估计也认为 speech recognition也是个非常简单的问题。
我不知道你是否清楚 state-of-the-art computer vision 的状况。
你能买到一个在1000以下reliable的simple stereo system吗？
为什么大家那么激动，因为以前想在一万以下打个fancy的computer vision的系统，基
本上是mission impossible。
看看hack这个东西的人是怎么说的吧。
http://www.memo.tv/kinect_why_it_matters
Kinect, being a depth sensing camera, gives us a ton of extra data over any
consumer device in it's price range. With that extra data, we are a lot more
knowledgable about what is happening in our environment, we can understand
it more accurately, thus we can create smarter systems that respond more
intuitively.
A lot of people are asking "what can you do with kinect that you couldn't
do before". Asking that question, is missing the point. It depends what
exactly "you" means. Is the question "What can I, Memo, do with kinect that
I couldn't do before?" Or is it "what could Myron Krueger do with kinect
that he couldn't before?" (answer is probably not much), or is it referring
to a more generic "you"?
Kinect is making nothing which wasn't already technically possible,
possible. It is just making it accessible, not just in terms of price, but
also in terms of simplicity and ease. The question should not be "what can
you do with kinect that you couldn't do before", but it should be "how much
simpler is it (technically) to do something with kinect, which was a lot
harder with consumer devices before kinect". To demonstrate what I mean,
here is a rough prototype I posted yesterday within a few hours of getting
my hands on a kinect.
Kinect is hooked up to my macbook pro, I'm using the opensource drivers
mentioned above to read the color image and depth map, and wrote the demo
prototype you see above. One hand draws in 3D, two hands rotates the view.
=========================================================================
Without kinect this is completely possible. You could use high end
expensive equipment, but you don't even need to. You could use two cheap
webcams, make sure you have good control of your lighting, you might need to
setup a few IR emitters, ideally try and get a clean unchanging background
(not essential but helps a lot). And then you will need a *lot* of hairy
maths, algorithms and code. I'm sure lots of people out there are thinking "
hey what's the big deal, I don't find those algorithms hairy at all, I could
do that without a Kinect, and I already have done". Well smartass this isn'
t about you.
With the kinect, you pretty much just plug it in, make sure there isn't any
bright sunlight around, and with a few lines of code you have the
information you need. You have that extra data that you can now use to do
whatever you want. Now that interaction is available for artists /
developers of *all* levels, not just the smelly geeks - and that is very
important. Once we have everybody designing, creating and playing with these
kinds of interactions - who prekinect would not have been able to - then we
will be smothered in amazing, innovative, fresh ideas and applications.
Sure we'll get thousands of pinch-to-zoom-and-rotate-the-photo demos, which
will get sickening pretty quickly, but amongst all that will be ideas that
you or I would have never thought of in a million years, but we'll instantly
fall in love with, and it will spark new ideas in us, sending us off in a
frenzy of creative development, which in turn feeds others and the cycle
continues.
And that's why it matters.
============================================================================
=====
【在 l****l 的大作中提到】
: simple binocular storeo. Why so many buzz...
|
