Regional Multi-scale Approach for Visually Pleasing Explanations of Deep Neural Networks
This work addresses the need for better interpretability in deep learning for users requiring clear visual explanations, though it is incremental as it builds on existing prediction difference methods.
The paper tackled the problem of generating class-discriminative and visually pleasing explanations for deep neural network predictions by proposing a region-based multi-scale approach, resulting in saliency maps that are much more effective in these aspects as demonstrated experimentally.
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.