2

i get solution to share convolution kernel by coping the kernels (share weights between keras layers but have other parameters differ). But in this solution, we just able to copy the kernels. How to share the kernels with kernel rotation in keras?

Matrixwira
  • 45
  • 1
  • 6
  • What do you mean by sharing rotation? – Jakub Bartczuk Apr 18 '18 at 12:59
  • 1
    let's say we have two layers. According to the solution linked above, we just able to copy **conv2.kernel = conv1.kernel**. But in my case, i want to flip 90 degree of the first layer's kernel and copy to the second layer's kernel. So, how to do like this in keras? **conv2.kernel = rotate(conv1.kernel)** – Matrixwira Apr 19 '18 at 03:02

0 Answers0