eduzhai > Applied Sciences > Engineering >

Axiom-based Grad-CAM Towards Accurate Visualization and Explanation of CNNs

  • king
  • (0) Download
  • 20210507
  • Save

... pages left unread,continue reading

Document pages: 18 pages

Abstract: To have a better understanding and usage of Convolution Neural Networks(CNNs), the visualization and interpretation of CNNs has attracted increasingattention in recent years. In particular, several Class Activation Mapping(CAM) methods have been proposed to discover the connection between CNN sdecision and image regions. In spite of the reasonable visualization, lack ofclear and sufficient theoretical support is the main limitation of thesemethods. In this paper, we introduce two axioms -- Conservation and Sensitivity-- to the visualization paradigm of the CAM methods. Meanwhile, a dedicatedAxiom-based Grad-CAM (XGrad-CAM) is proposed to satisfy these axioms as much aspossible. Experiments demonstrate that XGrad-CAM is an enhanced version ofGrad-CAM in terms of conservation and sensitivity. It is able to achieve bettervisualization performance than Grad-CAM, while also be class-discriminative andeasy-to-implement compared with Grad-CAM++ and Ablation-CAM. The code isavailable at this https URL.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×