Class Activation Map (CAM) or Saliency Maps for 1 dimensional data (vector)

[saanasum]

Is there a way in R-keras to generate CAMs or Saliency Maps for one-dimensional input (e.g. a vector) as described for example for the "keras-vis" package for python?: raghakot/keras-vis#93

I trained a CNN using R-Keras and now want to visualize where the CNN is focusing on when performing class predictions.

P.S.: Once I asked for the 2D option in R-Keras: #182 --> maybe the code proposed there is useful.

[crazyhottommy]

Hey, I have the exactly the same question. I want to calculate the salicency maps as in this example https://github.com/abidlabs/deep-learning-genomics-primer/blob/master/A_Primer_on_Deep_Learning_in_Genomics_Public.ipynb but with keras in R.