http://www.telegraph.co.uk/science/2018/...d-glasses/
There have been numerous attempts to give robots depth perceptions by using human-like stereoscopic two-eye vision. But that takes lots of processing power and it's slow. So researchers have been experimenting with Praying Mantises which also seem to possess depth perception but somehow do it very quickly with tiny little brains.
(Part of the experiments involved temporarily sticking stylish little shades to their heads with beeswax, making them look very cool.)
Photo here:
http://www.telegraph.co.uk/content/dam/s...mwidth=400
"Published in the journal Current Biology, the study concluded that the praying mantis vision system is "very robust" and simpler than that used by humans, meaning it could provide a far better template for robots."
From the journal:
"This is a fundamentally different stereo mechanism from that found in primates... Insect steropsis has thus evolved to be computationally efficient while being robust to poor image resolution and to discrepancies in the pattern of lumanance between the two eyes."
It seems to work by comparing differences in motion as seen by the two eyes, not by comparing perspective or foreground/background differences in static features of a still image seen by both eyes. So apparently our mantises with their cool shades only have depth perception for moving objects (like insect prey).
"Before the discovery of insect stereopsis, David Lee hypothesized that organisms whose "visual system were attuned to pick up primarily the kinetic structure of the optic array" might be able to use kinetic, but not static, disparity. The praying mantis seems to be a good example of such an organism. Mantis stereopsis is computationally simple enough to implement in a brain of one million neurons, and---remarkably--- successfully detects stereoscopic distance in images where human steropsis fails. This demonistrates that distinct evolutionary pressures can result in completely different algorithms for binocular stereopsis.
Journal article here:
http://www.cell.com/current-biology/full...18)30014-9
There have been numerous attempts to give robots depth perceptions by using human-like stereoscopic two-eye vision. But that takes lots of processing power and it's slow. So researchers have been experimenting with Praying Mantises which also seem to possess depth perception but somehow do it very quickly with tiny little brains.
(Part of the experiments involved temporarily sticking stylish little shades to their heads with beeswax, making them look very cool.)
Photo here:
http://www.telegraph.co.uk/content/dam/s...mwidth=400
"Published in the journal Current Biology, the study concluded that the praying mantis vision system is "very robust" and simpler than that used by humans, meaning it could provide a far better template for robots."
From the journal:
"This is a fundamentally different stereo mechanism from that found in primates... Insect steropsis has thus evolved to be computationally efficient while being robust to poor image resolution and to discrepancies in the pattern of lumanance between the two eyes."
It seems to work by comparing differences in motion as seen by the two eyes, not by comparing perspective or foreground/background differences in static features of a still image seen by both eyes. So apparently our mantises with their cool shades only have depth perception for moving objects (like insect prey).
"Before the discovery of insect stereopsis, David Lee hypothesized that organisms whose "visual system were attuned to pick up primarily the kinetic structure of the optic array" might be able to use kinetic, but not static, disparity. The praying mantis seems to be a good example of such an organism. Mantis stereopsis is computationally simple enough to implement in a brain of one million neurons, and---remarkably--- successfully detects stereoscopic distance in images where human steropsis fails. This demonistrates that distinct evolutionary pressures can result in completely different algorithms for binocular stereopsis.
Journal article here:
http://www.cell.com/current-biology/full...18)30014-9