Interpreting Low-level Vision Models with Causal Effect Maps

TL;DR

This paper introduces Causal Effect Maps (CEM), a causality-based interpretability method for low-level vision models, revealing insights about input-output relationships and challenging assumptions about input information and global mechanisms.

Contribution

The paper proposes a novel, model- and task-agnostic causality-based interpretability method called CEM for low-level vision models, providing new insights into their behavior.

Findings

More input information does not always improve performance.

Global receptive field mechanisms may be ineffective in denoising.

Multi-task training encourages local over global information use.

Abstract

Deep neural networks have significantly improved the performance of low-level vision tasks but also increased the difficulty of interpretability. A deep understanding of deep models is beneficial for both network design and practical reliability. To take up this challenge, we introduce causality theory to interpret low-level vision models and propose a model-/task-agnostic method called Causal Effect Map (CEM). With CEM, we can visualize and quantify the input-output relationships on either positive or negative effects. After analyzing various low-level vision tasks with CEM, we have reached several interesting insights, such as: (1) Using more information of input images (e.g., larger receptive field) does NOT always yield positive outcomes. (2) Attempting to incorporate mechanisms with a global receptive field (e.g., channel attention) into image denoising may prove futile. (3) Integrating multiple tasks to train a general model could encourage the network to prioritize local information over global context. Based on the causal effect theory, the proposed diagnostic tool can refresh our common knowledge and bring a deeper understanding of low-level vision models. Codes are available at https://github.com/J-FHu/CEM.