search

Tomoyon / makehuman

Automatically exported from code.google.com/p/makehuman
0 stars 0 forks source link

3d model of a real life person from a single image #108

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago 
an api and an interface for making motion of individual part will be a great 
feature. also some kind of intelligence (AI) code which will be capable of 
integrating with different image processing AIs so that to extract posing 
information from a single image of a person(or may be an interface to manual 
fit the posture of the model onto the image) with know light conditions like in 
sun (light coming from infinity) and the direction of light may be also be 
extracted from the same image. so that the image can be superimposed on a 
generic model of a person of known gender (again can be done from image 
processing AI) with the same posture that has been extracted from the image AI 
or given manually. now a really good correction algorithm for correcting the 
generic 3D model of the person superimposed over the image will all that will 
be needed to get the complete 3D model of a real life human being from one 
single image. with the texture information only for the visible parts.

Original issue reported on code.google.com by vender...@gmail.com on 17 May 2011 at 10:10

GoogleCodeExporter commented 9 years ago 
Unfortunately the AI that you're referring to here is called 'strong AI', which 
doesn't exist now and (depending on your position re: several philosohical 
arguments about intelligence) likely won't exist for a long time, if ever.

Original comment by damian.s...@gmail.com on 7 Dec 2012 at 8:00

GoogleCodeExporter commented 9 years ago 
Sorry, but this is not even remotely a "strong AI" problem, just an ordinary 
computer-vision problem. Movie studios use multiple cameras to do motion 
capture because it is better constrained, but pose-recovery from a single image 
is possible.

Rather than do a lot of literature searching to prove my point, I suggest any 
interested person google "human pose recovery from single image" and peruse a 
handful of the many results that come back. The top hit when I did it a few 
minutes ago was "Recovering 3D Human Pose from Monocular Images" by Agarwal and 
Triggs, 2004. You can look through the paper and see pictures of the poses they 
recover.

Original comment by frot...@gmail.com on 14 Apr 2013 at 1:49

GoogleCodeExporter commented 9 years ago

Original comment by mbasti...@gmail.com on 12 Jul 2013 at 9:34