CyCADA: Cycle-Consistent Adversarial Domain Adaptation

TLDR

Domain adaptation is essential for new environments, yet adversarial feature‑space models struggle with pixel‑level shifts, while recent GANs with cycle‑consistency can map images between domains without aligned pairs. The authors propose a discriminatively trained Cycle‑Consistent Adversarial Domain Adaptation (CyCADA) model. CyCADA adapts pixel‑ and feature‑level representations, enforces cycle‑consistency with a task loss, and requires no aligned pairs, enabling use across diverse visual recognition and prediction tasks. The model achieves state‑of‑the‑art results on digit classification and road‑scene semantic segmentation, successfully transferring from synthetic to real‑world domains.

Abstract

Domain adaptation is critical for success in new, unseen environments. Adversarial adaptation models applied in feature spaces discover domain invariant representations, but are difficult to visualize and sometimes fail to capture pixel-level and low-level domain shifts. Recent work has shown that generative adversarial networks combined with cycle-consistency constraints are surprisingly effective at mapping images between domains, even without the use of aligned image pairs. We propose a novel discriminatively-trained Cycle-Consistent Adversarial Domain Adaptation model. CyCADA adapts representations at both the pixel-level and feature-level, enforces cycle-consistency while leveraging a task loss, and does not require aligned pairs. Our model can be applied in a variety of visual recognition and prediction settings. We show new state-of-the-art results across multiple adaptation tasks, including digit classification and semantic segmentation of road scenes demonstrating transfer from synthetic to real world domains.