Uncertainty-aware Generalized Adaptive CycleGAN

TL;DR

This paper introduces UGAC, a probabilistic approach for unpaired image translation that models uncertainty and outliers, improving robustness and quality over existing methods in natural and medical imaging tasks.

Contribution

The paper presents UGAC, a novel probabilistic CycleGAN variant that explicitly models residual uncertainty with generalized Gaussian distributions, enhancing robustness to out-of-distribution data.

Findings

Superior image quality in denoising and modality translation tasks.

Enhanced robustness to out-of-distribution test data.

Outperforms state-of-the-art methods on quantitative metrics.

Abstract

Unpaired image-to-image translation refers to learning inter-image-domain mapping in an unsupervised manner. Existing methods often learn deterministic mappings without explicitly modelling the robustness to outliers or predictive uncertainty, leading to performance degradation when encountering unseen out-of-distribution (OOD) patterns at test time. To address this limitation, we propose a novel probabilistic method called Uncertainty-aware Generalized Adaptive Cycle Consistency (UGAC), which models the per-pixel residual by generalized Gaussian distribution, capable of modelling heavy-tailed distributions. We compare our model with a wide variety of state-of-the-art methods on two challenging tasks: unpaired image denoising in the natural image and unpaired modality prorogation in medical image domains. Experimental results demonstrate that our model offers superior image generation quality compared to recent methods in terms of quantitative metrics such as signal-to-noise ratio and structural similarity. Our model also exhibits stronger robustness towards OOD test data.