U-Noise: Learnable Noise Masks for Interpretable Image Segmentation

TL;DR

This paper presents U-Noise, a novel method for interpreting image segmentation models by learning noise masks that identify regions where noise can be added without affecting model performance, enabling quantitative interpretability.

Contribution

The paper introduces a learnable noise mask approach for interpretability in image segmentation, allowing quantitative evaluation based on downstream task performance.

Findings

The method effectively identifies regions where noise does not impact segmentation accuracy.

Qualitative comparison shows competitive interpretability with existing techniques.

Quantitative evaluation demonstrates the method's ability to assess feature importance.

Abstract

Deep Neural Networks (DNNs) are widely used for decision making in a myriad of critical applications, ranging from medical to societal and even judicial. Given the importance of these decisions, it is crucial for us to be able to interpret these models. We introduce a new method for interpreting image segmentation models by learning regions of images in which noise can be applied without hindering downstream model performance. We apply this method to segmentation of the pancreas in CT scans, and qualitatively compare the quality of the method to existing explainability techniques, such as Grad-CAM and occlusion sensitivity. Additionally we show that, unlike other methods, our interpretability model can be quantitatively evaluated based on the downstream performance over obscured images.