Society of Robots - Robot Forum
General Misc => Misc => Topic started by: Admin on June 15, 2006, 09:17:44 PM
-
Finally one of the unsolved holy grails in robotics has been solved.
It is perfectly normal for a person to look at a 2D picture and be able to reconstruct the scene in 3D. But to get a computer to do that had never been accomplished.
These guys from Carnegie Mellon University used machine learning to statistically figure out how to construct a 3D scene from just a 2D pic.
http://www.cmu.edu/PR/releases06/060613_3d.html
No more need for laser rangefinders! The next step of course is to put this on a robot . . .
-
Awesome...and the best part is that i am learning machine intelligence in my Masters Program ...dont know why i get so excited when learning new things... thanks for the share.. :D
-
Finally one of the unsolved holy grails in robotics has been solved.
Cool, but not unsolved or new. Stanford had a website doing this three years ago.
http://techcrunch.com/2008/02/27/make3d-turn-a-2d-picture-into-a-3d-model/ (http://techcrunch.com/2008/02/27/make3d-turn-a-2d-picture-into-a-3d-model/)
-
Cool, but not unsolved or new. Stanford had a website doing this three years ago.
This thread was originally posted on June 15, 2006, 11:17:44 PM
(almost 5 years ago)
:P
-
sorry for the mix up..by new i meant 'new-to-me'...cause a couple of days ago a friend and i were discussing face recognition problems, and if we could converge a 2d image to 3d for better feature extraction, and i said that something like this must've have been done already :D
and i stumbled upon this here at SOR..!
-
Cool, but not unsolved or new. Stanford had a website doing this three years ago.
This thread was originally posted on June 15, 2006, 11:17:44 PM
(almost 5 years ago)
:P
My bad....
-
I didn't give it any serious look, but how real time is it?
For it to be practical it needs real time modelling.
Processors are getting more and more serious about their power while lowering their power consumption.
But consumption and is very critical (plus with volume and weight)? I can't really run a robot with a
computer or power Grid following it.... ;-)
I don't try to be negative, just asking.
BTW, it's not the first 3D from camera, I've seen a 3D camera before, but from what I remember I used two CCD sensors,
not one.