Grad-CAM in Pytorch: Use of Forward and Backward Hooks

Using gradients to understand how your model predicts

I took notice of a technique called Grad-CAM that enables the inspection of how a convolutional neural network predicts its outputs. For example, in a classifier, you can gain insight into how your neural network used the input to make its prediction. It all started with the original paper that described it. In this article, we’re going to implement it …

author classifier convolutional-network convolutional neural network dataset deep learning example grad-cam image insight kaggle network neural network python pytorch ray x-ray