Guided Back-propagation with tensorflow backend

[prab...@gmail.com]

Hello folks,

Has anybody been able to implement guided back-propagation? 

I am trying to visualize what my model has learned. I am trying to use the grad-CAM paper. Keras-vis is an excellent library that implements grad-CAM but it lacks guided back-prop.

Last time this was asked was in Aug-Sept of 2016. And, it went unanswered. Just wondering if there was an update.

cheers,

Prabhat Garg

[ins...@gmail.com]

I also wanted to visualize my model for Kaggle competition. I needed to debug my model. So I made it for me. 

I hope it also works for your case. 

link: https://github.com/insikk/Grad-CAM-tensorflow

[prab...@gmail.com]

Thanks for the reply. I took a look at the link. The work looks very interesting. However, I think its primarily based in raw tensorflow. Is there a way for me to use a keras model?