A Comprehensive Survey on Test-Time Adaptation under Distribution Shifts

Abstract

Machine learning methods strive to acquire a robust model during the training process that can effectively generalize to test samples, even in the presence of distribution shifts. However, these methods often suffer from performance degradation due to unknown test distributions. Test-time adaptation (TTA), an emerging paradigm, has the potential to adapt a pre-trained model to unlabeled data during testing, before making predictions. Recent progress in this paradigm has highlighted the significant benefits of using unlabeled data to train self-adapted models prior to inference. In this survey, we categorize TTA into several distinct groups based on the form of test data, namely, test-time domain adaptation, test-time batch adaptation, and online test-time adaptation. For each category, we provide a comprehensive taxonomy of advanced algorithms and discuss various learning scenarios. Furthermore, we analyze relevant applications of TTA and discuss open challenges and promising areas for future research. For a comprehensive list of TTA methods, kindly refer to \url{https://github.com/tim-learn/awesome-test-time-adaptation}.

Discussions, comments, and questions are all welcomed in \url{https://github.com/tim-learn/awesome-test-time-adaptation}

References in corpus (91)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• A Simple Framework for Contrastive Learning of Visual Representations
• Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
• Learning Transferable Features with Deep Adaptation Networks
• Unsupervised Domain Adaptation by Backpropagation
• FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence
• Unsupervised Learning of Visual Features by Contrasting Cluster Assignments
• Unsupervised Data Augmentation for Consistency Training
• Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results
• Unsupervised Representation Learning by Predicting Image Rotations
• Domain Generalization: A Survey
• CyCADA: Cycle-Consistent Adversarial Domain Adaptation
• A review of domain adaptation without target labels
• MixMatch: A Holistic Approach to Semi-Supervised Learning
• Model Adaptation: Unsupervised Domain Adaptation without Source Data
• Countering Adversarial Images using Input Transformations
• Learning Discrete Representations via Information Maximizing Self-Augmented Training
• Tent: Fully Test-time Adaptation by Entropy Minimization
• Batch Renormalization: Towards Reducing Minibatch Dependence in Batch-Normalized Models
• Semantic Photo Manipulation with a Generative Image Prior
• Test-Time Adaptable Neural Networks for Robust Medical Image Segmentation
• Understanding Measures of Uncertainty for Adversarial Example Detection
• Improving robustness against common corruptions by covariate shift adaptation
• Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation
• Information-Theoretical Learning of Discriminative Clusters for Unsupervised Domain Adaptation
• Test-Time Prompt Tuning for Zero-Shot Generalization in Vision-Language Models
• Detecting and Correcting for Label Shift with Black Box Predictors
• Model Adaptation: Historical Contrastive Learning for Unsupervised Domain Adaptation without Source Data
• Evaluating Prediction-Time Batch Normalization for Robustness under Covariate Shift
• MEMO: Test Time Robustness via Adaptation and Augmentation
• ADATIME: A Benchmarking Suite for Domain Adaptation on Time Series Data
• Test-Time Training with Self-Supervision for Generalization under Distribution Shifts
• Generative Domain Adaptation for Face Anti-Spoofing
• Adversarial Style Mining for One-Shot Unsupervised Domain Adaptation
• Efficient Test-Time Model Adaptation without Forgetting
• Self-Supervised Policy Adaptation during Deployment
• Towards Stable Test-Time Adaptation in Dynamic Wild World
• S-Prompts Learning with Pre-trained Transformers: An Occam's Razor for Domain Incremental Learning
• Regularized Learning for Domain Adaptation under Label Shifts
• Unsupervised Domain Adaptation of Black-Box Source Models
• Adversarial purification with Score-based generative models
• NOTE: Robust Continual Test-time Adaptation Against Temporal Correlation
• Divide and Contrast: Source-free Domain Adaptation via Adaptive Contrastive Learning
• Adaptive Risk Minimization: Learning to Adapt to Domain Shift
• Balancing Discriminability and Transferability for Source-Free Domain Adaptation
• A Prototype-Oriented Framework for Unsupervised Domain Adaptation
• Source Data-absent Unsupervised Domain Adaptation through Hypothesis Transfer and Labeling Transfer
• Multi-step Online Unsupervised Domain Adaptation
• Confidence Score for Source-Free Unsupervised Domain Adaptation
• When Source-Free Domain Adaptation Meets Learning with Noisy Labels
• TTN: A Domain-Shift Aware Batch Normalization in Test-Time Adaptation
• Inference Stage Optimization for Cross-scenario 3D Human Pose Estimation
• Variational Model Perturbation for Source-Free Domain Adaptation
• Towards Inheritable Models for Open-Set Domain Adaptation
• Divide to Adapt: Mitigating Confirmation Bias for Domain Adaptation of Black-Box Predictors
• Align Your Prompts: Test-Time Prompting with Distribution Alignment for Zero-Shot Generalization
• Rethinking Precision of Pseudo Label: Test-Time Adaptation via Complementary Learning
• Test time Adaptation through Perturbation Robustness
• Test-Time Adaptation via Self-Training with Nearest Neighbor Information
• Online Adversarial Purification based on Self-Supervision
• Revisiting Realistic Test-Time Training: Sequential Inference and Adaptation by Anchored Clustering
• Empowering Graph Representation Learning with Test-Time Graph Transformation
• Unsupervised Domain Adaptation in the Absence of Source Data
• Test-Time Training Can Close the Natural Distribution Shift Performance Gap in Deep Learning Based Compressed Sensing
• Adaptive Domain Generalization via Online Disagreement Minimization
• Test-Time Robust Personalization for Federated Learning
• POUF: Prompt-oriented unsupervised fine-tuning for large pre-trained models
• If your data distribution shifts, use self-learning
• In Search for a Generalizable Method for Source Free Domain Adaptation
• Learning to Generalize across Domains on Single Test Samples
• Tailoring: encoding inductive biases by optimizing unsupervised objectives at prediction time
• Re-using Adversarial Mask Discriminators for Test-time Training under Distribution Shifts
• Optimization-Free Test-Time Adaptation for Cross-Person Activity Recognition
• Adaptive Test-Time Personalization for Federated Learning
• Test-Time Adaptation with CLIP Reward for Zero-Shot Generalization in Vision-Language Models
• C-TPT: Calibrated Test-Time Prompt Tuning for Vision-Language Models via Text Feature Dispersion
• Union-set Multi-source Model Adaptation for Semantic Segmentation
• Efficient Test-Time Adaptation for Super-Resolution with Second-Order Degradation and Reconstruction
• Open-World Pose Transfer via Sequential Test-Time Adaption
• Cluster-level pseudo-labelling for source-free cross-domain facial expression recognition
• Feed-Forward Latent Domain Adaptation
• Temporal Coherent Test-Time Optimization for Robust Video Classification
• RDumb: A simple approach that questions our progress in continual test-time adaptation
• Semantic Self-adaptation: Enhancing Generalization with a Single Sample
• Domain Alignment Meets Fully Test-Time Adaptation
• Energy-Based Test Sample Adaptation for Domain Generalization
• Towards Understanding GD with Hard and Conjugate Pseudo-labels for Test-Time Adaptation
• Controllable Style Transfer via Test-time Training of Implicit Neural Representation
• Active Test-Time Adaptation: Theoretical Analyses and An Algorithm
• AdaptGuard: Defending Against Universal Attacks for Model Adaptation
• Test-Time Training on Nearest Neighbors for Large Language Models

Cited by in corpus (6)

• Enhancing and Adapting in the Clinic: Source-free Unsupervised Domain Adaptation for Medical Image Enhancement
• T-TIME: Test-Time Information Maximization Ensemble for Plug-and-Play BCIs
• Multi-Stream Cellular Test-Time Adaptation of Real-Time Models Evolving in Dynamic Environments
• Analyzing Effects of Mixed Sample Data Augmentation on Model Interpretability
• Predict the Retrieval! Test time adaptation for Retrieval Augmented Generation
• Data-Efficient Stream-Based Active Distillation for Scalable Edge Model Deployment