How to know which features are selected from an image while classifying using transfer learning?

I am doing medical image classification using transfer learning. How to visualize which features or which part of the images are selected from training data to classify the image? Can I use CAM or Grad-CAM for this purpose? If Yes how?

Hi,

There are multiple methods that can be used.
I think the captum package is what you’re looking for !

1 Like