A light-weight method to foster the (Grad)CAM interpretability and explainability of classification networks

Alfred Schöttl

We consider a light-weight method which allows to improve the explainability of localized classification networks. The method considers (Grad)CAM maps during the training process b…