Improving the Image Kernels spreadsheet

I have found the image kernels spreadsheet to be very useful. I have made some minor changes to the spreadsheet.

  1. Now you can play with some of the kernels within the excel spreadsheet itself.
  2. Added visualisation of convolution layers. It shows how the calculation is done instead of looking at the formula. Right now its done for just one layer, but will be doing to across the whole spreadsheet.
  3. Added the MNIST dataset. You can pick any number from that dataset.

Any recommendation to develop this is further? Adding activation functions etc?


Edit: Github