Anyone know anything about vision ? How it works? Computational approaches to vision?
How we robots can recognise stuff?
Any fundamentals in vision?
>>8501204
DUDE PRIMES
>>8501202
>we robots
>we
>robots
OP, r u the tobor?
>>8501210
looool how can *
Read David Marr, although a lot of his work has become kind of obsolete.
Vision is becoming very well understood though, we have pretty good mappings of visual cortex.
We know that the basic approach the brain uses is hierarchical feature detection. Low levels of cortex detect basic shapes like lines and edges, higher levels detect more complex objects.
I think the new direction of the field is looking at vision as "inverse graphics". If you look at graphics as going from an entity model, to pixels on a screen, how do you go from pixels back to an entity model. Very fascinating ideas.
>>8501202
I know a little.
I use to work in a machine vision lab.
>>8501688
how weere you worjking there? what do you do now? i want someway to know contemporary/updated views of machine vision, how does it compare to the capabilities of our own vision, do they design.. i dont know what you call it... neural networks that are biologically plausibly related to real vision? are there any fundamental ideas or constraints that seem to be recurring across machine vision and techniques used?
>>8501672
im looking for a good textbook or something about visual processing from reitina to cortex and is relatively up to date in terms of generally accepted ideas. i know of david marr and his esteem but im not so interested in outdated ideas.
>>8501202
That depends, is the global maximum for your utility function at "kill all humans" or "do not kill all humans"?
>>8501202
Seems convolutional neural nets or variants of it seem to work best these days.