How to use three channel images in FeedForward Neural Network?

467 Views Asked by At

As FeedForward Neural Networks (Not CNN) has only one Layer of Data i.e. It has the ability to process only GrayScale images. How do we make it process Color (RGB) 3-Channel images?

1

There are 1 best solutions below

0
AudioBubble On BEST ANSWER

I would do the following methods.

  1. Add pixel-wise values of RGB channels and give it as 1-channel image to your Feedforward Neural Network.
  2. Add pixel-wise values of RGB channel, divide pixel values by 3 and give it as grey-scale image to your Feedforward Neural Network.
  3. Perpend 3 Linear layers to your model, apply a non-linear function, add 3 Layers results Neuron wise and then give it your as input to your Feedforward Neural Network.