From: raedomn@gmail.com
Hi Jeremy,
On Apr 29, 6:41 am, "ntlworld.com" wrote:
> Hi,
>
> What is the current state-of-the-art with respect to 'computer vision' and
> the recognition of 3-dimensional objects?
>
> I read somewhere that a bit of a holy grail at the moment would be to allow
> a computer to have the 3D object recognition of a 4yr old child? I'm
> surprised we're not near that yet? Is that true?
Yes. I would think this is mostly true. Human vision is amazing in
that even 4yr olds can recognize the same object from different
perspectives, yet it is inherently difficult to training a
computational model to perform the same task with comparable accuracy.
This is in fact a big problem in the field of computer vision.
>
> What is the basic problem with vision? Is it because the thing is so
> reliant on heavy processing? Would the field be improved by the increase in
> raw computing power or are we still missing fundamental understanding of the
> problem?
Yes and no. The first thing is that we do not understand too much
about human vision. We know the retina captures visual information and
pass it along through cortical and subcortical pathways to Primary
Visual Cortex. However, the complicated mechanism in this pathway, and
how different parts of visual cortex (such as V1, V2, and etc) analyze
this information is largely unknown. People generally feel it would
require too much computational resources if we simulate the brain (to
the extent what know today) faithfully. But without doing so, ad-hoc
strategies are inherently problem-specific and weak in achieving
general vision abilities like that of humans.
>
> Regards
> Jeremy Watts
>
Best,
Ting
[ comp.ai is moderated ... your article may take a while to appear. ]
--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
|
