Re:Ccv open sources near state-of-the-art image classifier u...
by bsaul on Thursday March 27, 2014 @11:49AM
(#7481088)
is there any image recognition software working with texturized 3D models as input for data training ?
[ Parent
| on HN ]
Re:Ccv open sources near state-of-the-art image classifier u...
by Qworg on Thursday March 27, 2014 @02:09PM
(#7482116)
I've seen systems that perform standard image recognition on one side and point cloud integration on the other, then combining them.
I don't know if they're open source though or even available.
Re:Ccv open sources near state-of-the-art image classifier u...
by frozenport on Friday March 28, 2014 @02:24AM
(#7485659)
Thats a wierd one, how would you rotate them? Can't you just extract them as a 2D texture and operate on that?
Re:Ccv open sources near state-of-the-art image classifier u...
by nitrogen on Thursday March 27, 2014 @12:35PM
(#7481490)
Can you treat Z as just another color channel and use the same algorithms?
Re:Ccv open sources near state-of-the-art image classifier u...
by kastnerkyle on Thursday March 27, 2014 @01:25PM
(#7481892)
It depends - some algorithms exploit correlations between pixels to reduce the computational load. This is bad if the Z channel is not strongly correlated with the others (RGB are very strongly correlated in natural images). Since depth shouldn't usually be correlated with color, this might cause some issues. Some experiments were done in "Learning Feature Representations with K-means", A. Ng. and A. Coates. (http://www.stanford.edu/~acoates/papers/coatesng_nntot2012.p...)
Give it a try, but also look at the underlying assumptions of your algorithm if it performs poorly.