Does computer vision intentionally mimic the vision of a human or is it just coincidental that a good computer vision system (with convolutional neural networks for example) reassembles some properties of the human vision apparatus?
$\begingroup$
$\endgroup$
0
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
I would argue that the similarities are largely superficial. More "classical" machine vision approaches focused on finding features a human can identify such as lines, texture and blobs of color.
Neural networks were originally designed based on what was known of neurons at the time, for example the activation functions are loosely based on how signals in the brain propagate. However, it is important not to take this analogy too far - there are many ways and nuances to how the brain operates and we are a long long way from being able to fully describe the brain.
While it's easy to anthropomorphize CNNs by saying things like "this layer pulls out background" or "this layer looks for edges", they are really just a bunch of stacked filters which are trained for a specific set of data. Ian Goodfellow has done some good work showing how these networks can be fooled by adding imperceptible (to a human) amounts of noise.
