How do I represent a color as an activation value within a neuron? ( might be off-topic)

I want to detect colors Using Neural Nets

In my Knowledge any activation function which is generally used will push the values in between 0-1 which isn't of any use to me...

Just wondering how Windows 10 decides on the theme color change when we switch themes.. (led me to think about this or its just simple averaging on pixel values of the channels?)

So seems like an impossible task for Neural Nets then?


1 Answer 1


Normally color spaces are not considered to be one dimensional.

Given three types of human cone cells, the most natural approach is probably not to use a single but three neurons. If color input is encoded as RGB values, then one neuron would be for red, one for green, and one for the blue channel.

  • $\begingroup$ But then is it still a neural net? That's more like A CNN as they are volume spaces $\endgroup$
    – Aditya
    Commented Mar 23, 2018 at 0:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.