Domain Generalization with MixStyle

2021/04/05 by Kaiyang Zhou, Zhou, Kaiyang, Yongxin Yang +5 · 43 citations
Computer Science · #Advanced Image and Video Retrieval Techniques #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Human Pose and Action Recognition #Machine Learning (cs.LG)

paper · pdf · doi:10.48550/arxiv.2104.02008

openalex publication_date 2021/04/05 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Though convolutional neural networks (CNNs) have demonstrated remarkable ability in learning discriminative features, they often generalize poorly to unseen domains. Domain generalization aims to address this problem by learning from a set of source domains a model that is generalizable to any unseen domain. In this paper, a novel approach is proposed based on probabilistically mixing instance-level feature statistics of training samples across source domains. Our method, termed MixStyle, is motivated by the observation that visual domain is closely related to image style (e.g., photo vs.~sketch images). Such style information is captured by the bottom layers of a CNN where our proposed style-mixing takes place. Mixing styles of training instances results in novel domains being synthesized implicitly, which increase the domain diversity of the source domains, and hence the generalizability of the trained model. MixStyle fits into mini-batch training perfectly and is extremely easy to implement. The effectiveness of MixStyle is demonstrated on a wide range of tasks including category classification, instance retrieval and reinforcement learning.

Citations

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Gradient-based learning applied to document recognition
• ImageNet classification with deep convolutional neural networks
• Instance Normalization: The Missing Ingredient for Fast Stylization
• A Learned Representation For Artistic Style
• IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
• DropBlock: A regularization method for convolutional networks
• CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features
• Unlabeled Samples Generated by GAN Improve the Person Re-identification Baseline in vitro
• Domain Randomization for Transferring Deep Neural Networks from\n Simulation to the Real World
• Domain Generalization via Model-Agnostic Learning of Semantic Features
• Random Erasing Data Augmentation

Cited by

• Towards Reliable Test-Time Adaptation: Style Invariance as a Correctness Likelihood
• Pathology-Aware Prototype Evolution via LLM-Driven Semantic Disambiguation for Multicenter Diabetic Retinopathy Diagnosis
• Generalizing Vision-Language Models with Dedicated Prompt Guidance
• RobustSurg: Tackling domain generalisation for out-of-distribution surgical scene segmentation
• Semantic-aware Random Convolution and Source Matching for Domain Generalization in Medical Image Segmentation
• Shoe Style-Invariant and Ground-Aware Learning for Dense Foot Contact Estimation
• AngioDG: Interpretable Channel-informed Feature-modulated Single-source Domain Generalization for Coronary Vessel Segmentation in X-ray Angiography
• A Dual-stage Prompt-driven Privacy-preserving Paradigm for Person Re-Identification
• Generalized Category Discovery under Domain Shift: A Frequency Domain Perspective
• ODP-Bench: Benchmarking Out-of-Distribution Performance Prediction
• Instance-Level Composed Image Retrieval
• DeepShield: Fortifying Deepfake Video Detection with Local and Global Forgery Analysis
• Exploring Scale Shift in Crowd Localization under the Context of Domain Generalization
• HIDISC: A Hyperbolic Framework for Domain Generalization with Generalized Category Discovery
• Connecting Domains and Contrasting Samples: A Ladder for Domain Generalization
• EReLiFM: Evidential Reliability-Aware Residual Flow Meta-Learning for Open-Set Domain Generalization under Noisy Labels
• Feature Stylization and Domain-aware Contrastive Learning for Domain Generalization
• High-Rate Mixout: Revisiting Mixout for Robust Domain Generalization
• EgoNight: Towards Egocentric Vision Understanding at Night with a Challenging Benchmark
• ResAD++: Towards Class Agnostic Anomaly Detection via Residual Feature Learning
• TRUST: Test-Time Refinement using Uncertainty-Guided SSM Traverses
• Shared Neural Space: Unified Precomputed Feature Encoding for Multi-Task and Cross Domain Vision
• Harnessing Synthetic Data from Generative AI for Statistical Inference
• Evaluating Synthetic Data Generation for Domain Generalization in Fetal Brain MRI Segmentation
• Self Identity Mapping
• SWAD: Domain Generalization by Seeking Flat Minima
• Preserving Domain Generalization in Fine-Tuning via Joint Parameter Selection
• Adaptive Knowledge Distillation using a Device-Aware Teacher for Low-Complexity Acoustic Scene Classification
• Vision-Language Semantic Aggregation Leveraging Foundation Model for Generalizable Medical Image Segmentation
• ConstStyle: Robust Domain Generalization with Unified Style Transformation
• Single Domain Generalization in Diabetic Retinopathy: A Neuro-Symbolic Learning Approach
• Quantization Meets OOD: Generalizable Quantization-aware Training from a Flatness Perspective
• Feed Two Birds with One Scone: Exploiting Function-Space Regularization for Both OOD Robustness and ID Fine-Tuning Performance
• Mix, Align, Distil: Reliable Cross-Domain Atypical Mitosis Classification
• Collaborative Semantic Aggregation and Calibration for Federated Domain Generalization
• Rethinking the Potential of Layer Freezing for Efficient DNN Training
• Revealing the Role of Audio Channels in ASR Performance Degradation
• Confounder Identification-free Causal Visual Feature Learning
• CoCAViT: Compact Vision Transformer with Robust Global Coordination
• SPA++: Generalized Graph Spectral Alignment for Versatile Domain Adaptation
• PatchDSU: Uncertainty Modeling for Out of Distribution Generalization in Keyword Spotting
• Multi-Granularity Feature Calibration via VFM for Domain Generalized Semantic Segmentation
• Learning Semantic Directions for Feature Augmentation in Domain-Generalized Medical Segmentation
• Rethink Domain Generalization in Heterogeneous Sequence MRI Segmentation

Related

• MixStyle Neural Networks for Domain Generalization and Adaptation
• CycleMix: Mixing Source Domains for Domain Generalization in Style-Dependent Data
• Frustratingly Simple Domain Generalization via Image Stylization
• FIXED: Frustratingly Easy Domain Generalization with Mixup
• Cross-Domain Ensemble Distillation for Domain Generalization