English

Interpretable and Fine-Grained Visual Explanations for Convolutional Neural Networks

Computer Vision and Pattern Recognition 2019-08-08 v1 Machine Learning

Abstract

To verify and validate networks, it is essential to gain insight into their decisions, limitations as well as possible shortcomings of training data. In this work, we propose a post-hoc, optimization based visual explanation method, which highlights the evidence in the input image for a specific prediction. Our approach is based on a novel technique to defend against adversarial evidence (i.e. faulty evidence due to artefacts) by filtering gradients during optimization. The defense does not depend on human-tuned parameters. It enables explanations which are both fine-grained and preserve the characteristics of images, such as edges and colors. The explanations are interpretable, suited for visualizing detailed evidence and can be tested as they are valid model inputs. We qualitatively and quantitatively evaluate our approach on a multitude of models and datasets.

Keywords

Cite

@article{arxiv.1908.02686,
  title  = {Interpretable and Fine-Grained Visual Explanations for Convolutional Neural Networks},
  author = {Jörg Wagner and Jan Mathias Köhler and Tobias Gindele and Leon Hetzel and Jakob Thaddäus Wiedemer and Sven Behnke},
  journal= {arXiv preprint arXiv:1908.02686},
  year   = {2019}
}

Comments

In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA, June 2019

R2 v1 2026-06-23T10:42:12.272Z