Abstract
Recently, many methods to interpret and visualize deep neural network predictions have been proposed and significant progress has been made. However, a more class-discriminative and visually pleasing explanation is required. Thus, this paper proposes a region-based approach that estimates feature importance in terms of appropriately segmented regions. By fusing the saliency maps generated from multi-scale segmentations, a more class-discriminative and visually pleasing map is obtained. We incorporate this regional multi-scale concept into a prediction difference method that is model-agnostic. An input image is segmented in several scales using the super-pixel method, and exclusion of a region is simulated by sampling a normal distribution constructed using the boundary prior. The experimental results demonstrate that the regional multi-scale method produces much more class-discriminative and visually pleasing saliency maps.
Highlights
Deep learning has facilitated breakthroughs for a variety of AI tasks and, in many cases, has achieved performance equal or superior to human performance [1]
This paper has proposed a new requirement—i.e., being visually pleasing—that is important for interpreting predictions in a perceptually attractive manner
This paper has presented the regional multi-scale prediction difference method as a viable solution
Summary
Deep learning has facilitated breakthroughs for a variety of AI tasks and, in many cases, has achieved performance equal or superior to human performance [1]. Most learning models do not satisfactorily explain why they reach a decision; model deployment is delayed or even abandoned Recognizing that this deficiency could cause potential harm, the European Union adopted a regulation for algorithmic decision-making that addresses the ‘‘right to explanation’’ [2]. Sensitivity analysis (SA) [7], the gradient-weighted class activation map (Grad-CAM) [5], and the proposed method are class-discriminative, whereas the other methods are not. This paper proposes another requirement, i.e., a visually pleasing saliency map. The results demonstrate that the proposed method produces much more class-discriminative and visually pleasing saliency maps (Fig. 1).
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
