0

I have an issue that seems to have no straight forward solution in Keras. My server runs on ubuntu 14.04, keras with backend tensorflow.

Here's the issue:

I have used keras.layers.Add() to add two outputs from two Conv2D layers in keras. Now I wish to invert this operation for a model. Any guidelines on how this can be implemented in Keras?

Axeman
  • 32,068
  • 8
  • 81
  • 94
geek
  • 65
  • 2
  • 11

1 Answers1

0

I think you should use the Functional API. It lets you create a network with multiple input/output.

So basically you need to merge the two conv outputs into one layer. In this way you will be able to track and invert the merge operation.

The link contains example implementations. If you provide the code that you are working on, I can update the answer to reflect the functional API version.

Vivek Kalyanarangan
  • 8,951
  • 1
  • 23
  • 42
  • 1
    The idea to track the outputs into the merge layer worked for me while inverting!! Thanks – geek Jan 15 '18 at 06:17