Robustness via Uncertainty-aware Cycle Consistency

Abstract

Unpaired image-to-image translation refers to learning inter-image-domain mapping without corresponding image pairs. Existing methods learn deterministic mappings without explicitly modelling the robustness to outliers or predictive uncertainty, leading to performance degradation when encountering unseen perturbations at test time. To address this, we propose a novel probabilistic method based on 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 various challenging tasks including unpaired image translation of natural images, using standard datasets, spanning autonomous driving, maps, facades, and also in medical imaging domain consisting of MRI. Experimental results demonstrate that our method exhibits stronger robustness towards unseen perturbations in test data. Code is released here: https://github.com/ExplainableML/UncertaintyAwareCycleConsistency.

arXiv commentAccepted at NeurIPS 2021. Code is at https://github.com/ExplainableML/UncertaintyAwareCycleConsistency. arXiv admin note: substantial text overlap with arXiv:2102.11747

References in corpus (6)

• What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?
• The Cityscapes Dataset for Semantic Urban Scene Understanding
• Aleatoric uncertainty estimation with test-time augmentation for medical image segmentation with convolutional neural networks
• Exploiting Epistemic Uncertainty of Anatomy Segmentation for Anomaly Detection in Retinal OCT
• Adversarial Self-Defense for Cycle-Consistent GANs
• DeepI2I: Enabling Deep Hierarchical Image-to-Image Translation by Transferring from GANs