mlsanityData to 7 Oct 2026

LEAD: Learning Decomposition for Source-free Universal Domain Adaptation

arXiv 2403.03421 ↗First posted 6 Mar 2024cs.CV cs.AI cs.LGcited by 25

Not on the map. Why?

Abstract (arXiv)

Universal Domain Adaptation (UniDA) targets knowledge transfer in the presence of both covariate and label shifts. Recently, Source-free Universal Domain Adaptation (SF-UniDA) has emerged to achieve UniDA without access to source data, which tends to be more practical due to data protection policies. The main challenge lies in determining whether covariate-shifted samples belong to target-private unknown categories. Existing methods tackle this either through hand-crafted thresholding or by developing time-consuming iterative clustering strategies. In this paper, we propose a new idea of LEArning Decomposition (LEAD), which decouples features into source-known and -unknown components to identify target-private data. Technically, LEAD initially leverages the orthogonal decomposition analysis for feature decomposition. Then, LEAD builds instance-level decision boundaries to adaptively identify target-private data. Extensive experiments across various UniDA scenarios have demonstrated the effectiveness and superiority of LEAD. Notably, in the OPDA scenario on VisDA dataset, LEAD outperforms GLC by 3.5% overall H-score and reduces 75% time to derive pseudo-labeling decision boundaries. Besides, LEAD is also appealing in that it is complementary to most existing methods. The code is available at https://github.com/ispc-lab/LEAD.

Key facts. Cited by 25 papers in the map (data through 7 Oct 2026). Builds on 48 of its 69 references that are in the map. First posted on arXiv 6 Mar 2024.

Builds on

48 of 69 references.

References · 48

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
  3. 2015Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networkscited by 9,617
  4. 2023Segment Anythingcited by 8,158
  5. 2016The Cityscapes Dataset for Semantic Urban Scene Understandingcited by 4,427
  6. 2015Domain-Adversarial Training of Neural Networkscited by 3,023
  7. 2016A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networkscited by 2,362
  8. 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,430
  9. 2018Moment Matching for Multi-Source Domain Adaptationcited by 1,327
  10. 2017Deep Hashing Network for Unsupervised Domain Adaptationcited by 1,181
  11. 2017CyCADA: Cycle-Consistent Adversarial Domain Adaptationcited by 1,079
  12. 2017Maximum Classifier Discrepancy for Unsupervised Domain Adaptationcited by 682
36 more
  1. 2018Recognition in Terra Incognitacited by 655
  2. 2015Optimal Transport for Domain Adaptationcited by 457
  3. 2018Domain Adaptive Faster R-CNN for Object Detection in the Wildcited by 442
  4. 2017VisDA: The Visual Domain Adaptation Challengecited by 432
  5. 2020Generalized ODIN: Detecting Out-of-distribution Image without Learning from Out-of-distribution Datacited by 408
  6. 2019High Frequency Component Helps Explain the Generalization of Convolutional Neural Networkscited by 310
  7. 2022Mitigating Neural Network Overconfidence with Logit Normalizationcited by 251
  8. 2020DACS: Domain Adaptation via Cross-domain Mixed Samplingcited by 226
  9. 2020Universal Source-Free Domain Adaptationcited by 219
  10. 2021Exploiting the Intrinsic Neighborhood Structure for Source-free Domain Adaptationcited by 176
  11. 2021Generalized Source-free Domain Adaptationcited by 175
  12. 2020Universal Domain Adaptation through Self Supervisioncited by 172
  13. 2019A review of domain adaptation without target labelscited by 154
  14. 2018Open Set Domain Adaptation by Backpropagationcited by 150
  15. 2019Progressive Domain Adaptation for Object Detectioncited by 134
  16. 2020Efficient Domain Generalization via Common-Specific Low-Rank Decompositioncited by 134
  17. 2020Learning to Balance Specificity and Invariance for In and Out of Domain Generalizationcited by 128
  18. 2017No More Discrimination: Cross City Adaptation of Road Scene Segmenterscited by 110
  19. 2022POEM: Out-of-Distribution Detection with Posterior Samplingcited by 108
  20. 2021Semantically Coherent Out-of-Distribution Detectioncited by 107
  21. 2021OVANet: One-vs-All Network for Universal Domain Adaptationcited by 92
  22. 2019Learning to Transfer Examples for Partial Domain Adaptationcited by 86
  23. 2022Training OOD Detectors in their Natural Habitatscited by 86
  24. 2017Partial Transfer Learning with Selective Adversarial Networkscited by 84
  25. 2021VDM-DA: Virtual Domain Modeling for Source Data-free Domain Adaptationcited by 69
  26. 2018Unsupervised Cross-Modality Domain Adaptation of ConvNets for Biomedical Image Segmentations with Adversarial Losscited by 60
  27. 2022Unified Optimal Transport Framework for Universal Domain Adaptationcited by 57
  28. 2022Divide and Contrast: Source-free Domain Adaptation via Adaptive Contrastive Learningcited by 53
  29. 2022BMD: A General Class-balanced Multicentric Dynamic Prototype Strategy for Source-free Domain Adaptationcited by 39
  30. 2023Upcycling Models under Domain and Category Shiftcited by 29
  31. 2021UMAD: Universal Model Adaptation under Domain and Category Shiftcited by 23
  32. 2022Prior Knowledge Guided Unsupervised Domain Adaptationcited by 20
  33. 2023Modality-Agnostic Debiasing for Single Domain Generalizationcited by 14
  34. 2023DEJA VU: Continual Model Generalization For Unseen Domainscited by 13
  35. 2020Open-Set Hypothesis Transfer with Semantic Consistencycited by 11
  36. 2021Mining Data Impressions from Deep Models as Substitute for the Unavailable Training Datacited by 7

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 9; 2025: 11; 2026: 5.

25 papers.

Sort:

By area:

  1. 2024GenMix: Effective Data Augmentation with Generative Diffusion Model Image Editingcited by 14
  2. 2024Multi-Granularity Class Prototype Topology Distillation for Class-Incremental Source-Free Unsupervised Domain Adaptationcited by 10
  3. 2026USBD: Universal Structural Basis Distillation for Source-Free Graph Domain Adaptationcited by 9
  4. 2024HGL: Hierarchical Geometry Learning for Test-time Adaptation in 3D Point Cloud Segmentationcited by 6
  5. 2024Recall and Refine: A Simple but Effective Source-free Open-set Domain Adaptation Frameworkcited by 6
  6. 2024MAP: MAsk-Pruning for Source-Free Model Intellectual Property Protectioncited by 4
  7. 2024GLC++: Source-Free Universal Domain Adaptation through Global-Local Clustering and Contrastive Affinity Learningcited by 4
  8. 2024Memory-Efficient Pseudo-Labeling for Online Source-Free Universal Domain Adaptation using a Gaussian Mixture Modelcited by 4
  9. 2025Target Semantics Clustering via Text Representations for Robust Universal Domain Adaptationcited by 4
  10. 2025Exploring Probabilistic Modeling Beyond Domain Generalization for Semantic Segmentationcited by 3
  11. 2024Tackling Dimensional Collapse toward Comprehensive Universal Domain Adaptationcited by 1
  12. 2025Training-Free Label Space Alignment for Universal Domain Adaptationcited by 1
13 more
  1. 2024Domain-invariant Representation Learning via Segment Anything Model for Blood Cell Classificationcited by 0
  2. 2025Analysis of Pseudo-Labeling for Online Source-Free Universal Domain Adaptationcited by 0
  3. 2025Beyond Retraining: Training-Free Unknown Class Filtering for Source-Free Open Set Domain Adaptation of Vision-Language Modelscited by 0
  4. 2025From Entanglement to Alignment: Representation Space Decomposition for Unsupervised Time Series Domain Adaptationcited by 0
  5. 2025Frustratingly Easy Feature Reconstruction for Out-of-Distribution Detectioncited by 0
  6. 2025Cross-Domain Attribute Alignment with CLIP: A Rehearsal-Free Approach for Class-Incremental Unsupervised Domain Adaptationcited by 0
  7. 2025Vicinity-Guided Discriminative Latent Diffusion for Privacy-Preserving Domain Adaptationcited by 0
  8. 2025Modality-Collaborative Low-Rank Decomposers for Few-Shot Video Domain Adaptationcited by 0
  9. 2025Autonomous Source Knowledge Selection in Multi-Domain Adaptationcited by 0
  10. 2026Safe-Subspace Pseudo-Label Refinement for Source-Free Graph Domain Adaptationcited by 0
  11. 2026LFM: Leveraging Foundation Models for Source-Free Universal Domain Adaptationcited by 0
  12. 2026Reliable Neural Collapse Approximation for Open-World Test-Time Adaptationcited by 0
  13. 2026A Comprehensive Benchmark of Source-Free Universal Domain Adaptation on Time Series Representationscited by 0