Re:4D toys
by gumby on Saturday August 04, 2018 @10:57AM
(#17687072)
If you start from the understandings (such as we have)of physiology and evolutionary psychology I think the answer is "no" (perhaps there's a trick someone can figure out via a different path -- I hope so).
It seems pretty clear that looking at 2D images (and in particular, still ones) is a learned task, like reading, rather than an innate one. Both appear to tie into deep structures in the brain, but both are very recent inventions.
From the perspective of developmental stages described by Piaget, children learn to view in three space primarily with objects within reach (parallax pretty much peters out around the ends of your arms. Once the child becomes mobile, she is able to use semantic understanding to estimate the size of distant objects and get a rough idea of distance. The whole human process of seeing is very different from the way, say, a NN is trained on an image: the whole thing isn't gulped in at once, but we foveate on various parts of the image and assemble / confabulate a whole. You can see this in the structure of Chinese classical painting or pre-persepctive European paintings: distant objects aren't sized in any way proportional to their apparent size. This really maps more to how much attention you pay to various objects in the scene.
You then learn to map that into a 3D model which I believe (but am not digging up refs this instance, sorry) has hardware support.
Thus the 2D->3D process exploits a lot of learned and innate knowledge and technique that you have already developed. With one exception you haven't any 4D experience. That one exception is temporal data -- we can easily extrapolate from, say, a sphere shrinking and growing. Apart from that, there isn't much to work with.
It seems pretty clear that looking at 2D images (and in particular, still ones) is a learned task, like reading, rather than an innate one. Both appear to tie into deep structures in the brain, but both are very recent inventions.
From the perspective of developmental stages described by Piaget, children learn to view in three space primarily with objects within reach (parallax pretty much peters out around the ends of your arms. Once the child becomes mobile, she is able to use semantic understanding to estimate the size of distant objects and get a rough idea of distance. The whole human process of seeing is very different from the way, say, a NN is trained on an image: the whole thing isn't gulped in at once, but we foveate on various parts of the image and assemble / confabulate a whole. You can see this in the structure of Chinese classical painting or pre-persepctive European paintings: distant objects aren't sized in any way proportional to their apparent size. This really maps more to how much attention you pay to various objects in the scene.
You then learn to map that into a 3D model which I believe (but am not digging up refs this instance, sorry) has hardware support.
Thus the 2D->3D process exploits a lot of learned and innate knowledge and technique that you have already developed. With one exception you haven't any 4D experience. That one exception is temporal data -- we can easily extrapolate from, say, a sphere shrinking and growing. Apart from that, there isn't much to work with.