beginning 3d camera, from colours only.

[rouncer]

Uses scale invarience,  but not to compare relative depths,   it uses it to normalize a 3d compute coming from a blurred sobel filter,  and i get this: 

sorta pops in and out at times, a 3dishness.
https://www.youtube.com/watch?v=0TmkVkKS5MQ


You can do alot better than this but - i just need a bit of time.
My new version of this is on the way,  with isotropic averaging and the relative depth comparison, that I was missing before.
I remember someone else was talking about 3d ages ago and said to use some kind of scale detector, which you can get from counting edge density.

[mklrobo]

I do not know what it is you are trying to accomplish. I
think I see a movie coming in and out, in a slanted, pixilated,
scale challenged, window.
If you could define more about what you are doing, maybe I
could offer a better opinion. If you are trying to get a robot
to see depth, maybe Watson's Deep mind SDK can help. 

[rouncer]

I bet watsons deep mind doesnt have anything to do with what I want,  if someone could convert rgb to 3d thats a HUGE gain, its not out yet.

Its rgb to a parallaxy type 3d effect,  using a pattern memory starting from rgb only,  maybe i havent quite got the symantics i need to do it.

I got my utube account deleted, so the video is not there anymore.   

I actually have videos of another implementation that doesnt go as good as this one did,  but thats going to change soon. (im going to get it working.)


Here, ill put it here but its not even as good as that war one,  cause I couldnt quite get the factors right.
https://www.youtube.com/watch?v=YnFdMnBGweg  <-in fact this one probably isnt working that well,  but i might as well put it here.