Deeply Explain CNN via Hierarchical Decomposition

Ming-Ming Cheng¹*, Peng-Tao Jiang¹*, Ling-Hao Han¹, Liang Wang², Philip Torr³

1-TMCC, Nankai University, 2-NLPR, 3-University of Oxford

Introduction

In computer vision, some attribution methods for explaining CNNs attempt to study how the intermediate features affect network prediction. However, they usually ignore the feature hierarchies among the intermediate features. This paper introduces a hierarchical decomposition framework to explain CNN’s decision-making process in a top-down manner. Specifically, we propose a gradient-based activation propagation (gAP) module that can decompose any intermediate CNN decision to its lower layers and find the supporting features. Then we utilize the gAP module to iteratively decompose the network decision to the supporting evidence from different CNN layers. The proposed framework can generate a deep hierarchy of strongly associated supporting evidence for the network decision, which provides insight into the decision-making process. Moreover, gAP is effort-free for understanding CNN-based models without network architecture modification and an extra training process. Experiments show the effectiveness of the proposed method. Please refer to our online demo.

If you find our paper is useful for your research, please consider citing:

@article{23IJCV-hDecomp,
  title={Deeply Explain CNN via Hierarchical Decomposition},
  author={Ming-Ming Cheng and Peng-Tao Jiang and Ling-Hao Han and Liang Wang and Philip Torr},
  journal={International Journal of Computer Vision},
  pages={–},
  volume={},
  year={2023},
  publisher={Springer}
}