CyCADA: Cycle-Consistent Adversarial Domain Adaptation

2017/11/08 by Judy Hoffman, Hoffman, Judy, Eric Tzeng +13 · 48 citations
Computer Science · #Anomaly Detection Techniques and Applications #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis

paper · pdf · doi:10.48550/arxiv.1711.03213

openalex publication_date 2017/11/08 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

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.

Citations

• Deep Domain Confusion: Maximizing for Domain Invariance
• Gradient-based learning applied to document recognition
• Adversarial Feature Learning
• Deep CORAL: Correlation Alignment for Deep Domain Adaptation
• Energy-based Generative Adversarial Network
• Learning to Generate Images of Outdoor Scenes from Attributes and\n Semantic Layouts
• Adversarial Discriminative Domain Adaptation
• Generative Visual Manipulation on the Natural Image Manifold
• Scribbler: Controlling Deep Image Synthesis with Sketch and Color

Cited by

• Source-Free Controlled Adaptation of Teachers for Continual Test-Time Adaptation
• Domain Adaptation in Structural Health Monitoring of Civil Infrastructure: A Systematic Review
• DistillFSS: Synthesizing Few-Shot Knowledge into a Lightweight Segmentation Model
• Do We Need Perfect Data? Leveraging Noise for Domain Generalized Segmentation
• Benchmarking Nighttime Traffic Sign Recognition with Illumination-Adaptive Detection and Semantic Attribute Reasoning
• FedSDA: Federated Stain Distribution Alignment for Non-IID Histopathological Image Classification
• Navigating the Wild: Pareto-Optimal Visual Decision-Making in Image Space
• Saliency-Guided Domain Adaptation for Left-Hand Driving in Autonomous Steering
• Domain Generalization for Semantic Segmentation: A Survey
• Instance-Aware Pseudo-Labeling and Class-Focused Contrastive Learning for Weakly Supervised Domain Adaptive Segmentation of Electron Microscopy
• Unsupervised Domain Adaptation via Similarity-based Prototypes for Cross-Modality Segmentation
• Curvilinear Structure-preserving Unpaired Cross-domain Medical Image Translation
• ST3D: Self-training for Unsupervised Domain Adaptation on 3D Object Detection
• A Framework for Low-Effort Training Data Generation for Urban Semantic Segmentation
• Uncertainty-Aware ControlNet: Bridging Domain Gaps with Synthetic Image Generation
• FRIEREN: Federated Learning with Vision-Language Regularization for Segmentation
• VirDA: Reusing Backbone for Unsupervised Domain Adaptation with Visual Reprogramming
• On Learning Invariant Representation for Domain Adaptation
• Signal Classification Recovery Across Domains Using Unsupervised Domain Adaptation
• Creative synthesis of kinematic mechanisms
• Bézier Meets Diffusion: Robust Generation Across Domains for Medical Image Segmentation
• Multilevel Knowledge Transfer for Cross-Domain Object Detection
• Shared Neural Space: Unified Precomputed Feature Encoding for Multi-Task and Cross Domain Vision
• Learning to see across Domains and Modalities
• Unsupervised Domain Adaptation for Object Detection via Cross-Domain Semi-Supervised Learning
• CCQA: Generating Question from Solution Can Improve Inference-Time Reasoning in SLMs
• Masked Feature Modeling Enhances Adaptive Segmentation
• Adversarial Appearance Learning in Augmented Cityscapes for Pedestrian Recognition in Autonomous Driving
• ImMimic: Cross-Domain Imitation from Human Videos via Mapping and Interpolation
• ACE and Diverse Generalization via Selective Disagreement
• Generalizable Model-agnostic Semantic Segmentation via Target-specific Normalization
• Delving into Robust Object Detection from Unmanned Aerial Vehicles: A Deep Nuisance Disentanglement Approach
• Generative AI for Enhanced Wildfire Detection: Bridging the Synthetic-Real Domain Gap
• IntrinsicReal: Adapting IntrinsicAnything from Synthetic to Real Objects
• Make me an Expert: Distilling from Generalist Black-Box Models into Specialized Models for Semantic Segmentation
• Significance-aware Information Bottleneck for Domain Adaptive Semantic Segmentation
• FOCUS: Frequency-Optimized Conditioning of DiffUSion Models for mitigating catastrophic forgetting during Test-Time Adaptation
• FixBi: Bridging Domain Spaces for Unsupervised Domain Adaptation
• Complete & Label: A Domain Adaptation Approach to Semantic Segmentation of LiDAR Point Clouds
• Domain Adaptation and Image Classification via Deep Conditional Adaptation Network
• Towards Shape Biased Unsupervised Representation Learning for Domain Generalization
• Diverse Teaching and Label Propagation for Generic Semi-Supervised Medical Image Segmentation
• Hierarchical Adaptive networks with Task vectors for Test-Time Adaptation
• ContourDiff: Unpaired Medical Image Translation with Structural Consistency
• TRUST: Leveraging Text Robustness for Unsupervised Domain Adaptation
• CORE-ReID V2: Advancing the Domain Adaptation for Object Re-Identification with Optimized Training and Ensemble Fusion
• Large-Scale Optimal Transport via Adversarial Training with Cycle-Consistency
• A Study into Echocardiography View Conversion

Related

• CrDoCo: Pixel-level Domain Transfer with Cross-Domain Consistency
• Gotta Adapt 'Em All: Joint Pixel and Feature-Level Domain Adaptation for Recognition in the Wild
• Generate To Adapt: Aligning Domains using Generative Adversarial Networks
• Unsupervised Pixel-Level Domain Adaptation with Generative Adversarial Networks
• Learning Domain Adaptive Features with Unlabeled Domain Bridges