TSJS-Grad-CAM

An implementation of Grad-CAM for tensorflowJS.
See the paper: https://arxiv.org/pdf/1610.02391v1.pdf
Gradient class activation maps (Grad-CAM) is a technique to visualize the decisions of deep learning networks.

Acknowledgement

This tensorflow-js implementation is based on the following repositories:

GradCAM implementation in keras:

https://github.com/jacobgil/keras-grad-cam

Tensorflow-js implementation of the CAM algorithm:

https://github.com/tensorflow/tfjs-examples/tree/master/visualize-convnet

Run Grad-CAM

The 'runGradCAM' file train a model on a pecified dataset and computes the grad-CAM for a given image.