pith. machine review for the scientific record. sign in

arxiv: 2604.17748 · v1 · submitted 2026-04-20 · 💻 cs.CV

Recognition: unknown

Source-Free Domain Adaptation with Vision-Language Prior

Song Tang , Yunxiang Bai , Wenxin Su , Mao Ye , Jianwei Zhang , Xiatian Zhu
Authors on Pith no claims yet

Pith reviewed 2026-05-10 05:27 UTC · model grok-4.3

classification 💻 cs.CV
keywords source-free domain adaptationvision-language modelsprompt learningmutual information maximizationknowledge distillationgap region reductionCLIP
0
0 comments X

The pith

A method called DIFO++ adapts models to new visual domains without any source data by customizing vision-language models like CLIP and distilling their knowledge into the target model.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Source-free domain adaptation tries to update a pre-trained model using only unlabeled data from a new target domain. Standard approaches depend on pseudo-labels that often contain errors. This paper shows that off-the-shelf vision-language models hold useful but generic knowledge that can be made task-specific by alternating two steps: first maximizing mutual information between the vision-language model and the target model through prompt learning, then distilling the customized knowledge back while focusing on the gap region of entangled and ambiguous features. Reliable pseudo-labels arise from fusing the two models' predictions with a memory buffer, and semantic alignment is enforced through category attention and consistency checks. If this works, it offers a way to reduce error accumulation in domain shifts such as medical imaging or scene recognition.

Core claim

Direct zero-shot use of a vision-language model on the target domain is unsatisfactory because it remains too generic. DIFO++ therefore alternates between customizing the vision-language model by maximizing mutual information with the target model in a prompt-learning manner and distilling the resulting knowledge into the target model by centering on gap-region reduction. During this process, the method identifies gap regions where features are entangled and class-ambiguous, generates reliable pseudo-labels by fusing predictions from both models with a memory mechanism, and guides alignment via category attention, predictive consistency, and referenced entropy minimization.

What carries the argument

The DIFO++ alternation of vision-language model customization (via mutual information maximization in prompt learning) and targeted knowledge distillation (via gap region reduction with fused pseudo-labels and category attention).

If this is right

  • Fusing predictions from the target model and the customized vision-language model with a memory buffer produces more reliable pseudo-labels than either model alone.
  • Focusing distillation on the gap region of entangled features captures richer task-specific semantics and improves semantic alignment.
  • Category attention combined with predictive consistency and referenced entropy minimization progressively reduces uncertainty during adaptation.
  • The two-step alternation yields measurable gains over prior source-free methods on standard benchmarks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same mutual-information customization step could be applied to other multimodal models beyond CLIP to supply priors for different adaptation tasks.
  • Gap-region identification might serve as a general diagnostic for where current domain-adaptation methods are most uncertain.
  • Extending the memory mechanism to longer adaptation sequences could further stabilize pseudo-label quality in streaming target data.

Load-bearing premise

Off-the-shelf vision-language models contain sufficiently rich and task-relevant knowledge that can be effectively customized for a specific target domain solely through mutual information maximization with the target model, without source data.

What would settle it

On a target domain whose classes have no overlap with the vision-language model's pre-training distribution, measure whether DIFO++ still outperforms standard source-free baselines that do not use the vision-language model.

Figures

Figures reproduced from arXiv: 2604.17748 by Jianwei Zhang, Mao Ye, Song Tang, Wenxin Su, Xiatian Zhu, Yunxiang Bai.

Figure 1
Figure 1. Figure 1: We expand beyond traditional SFDA methods [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of DIFO++: Our method alternates between two steps. First, we perform (a) [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: An illustrative idea of gap region reduction. (a) The gap region refers to the transitional area between class-consistent clusters, detected by the proposed met￾ric of referenced entropy. (b) Through a gap region￾driven knowledge adaptation, features within the gap region are properly consolidated, reducing its extent. (c) A snapshot of this progressive adaptation, in which Lrc guides the movement of featu… view at source ↗
Figure 4
Figure 4. Figure 4: Illustration of the design of category attention [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The performance of the scheme directly weighting [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The evolving dynamics of MMD distance during [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Convergence curves of different uncertainty met￾rics on the Ar→Cl task in Office-Home. Additionally, removing any loss component leads to per￾formance deterioration (rows 6–11). When Lcac or Lpc are unavailable, the average accuracy drops the most (by about 9%). The phenomenon is consistent with the cases in rows 3-4: as Lcac or Lpc work alone, the accu￾racy significantly improves. This observation confirm… view at source ↗
Figure 9
Figure 9. Figure 9: Feature distribution visualization on transfer task Ar [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Sensitivity of the hyper-parameters α, β, η, δ, γ, ϵ, and τ . 1 2 3 4 5 N 71.0 71.5 72.0 72.5 73.0 73.5 Accuracy (%) Office-Home (Ar->Cl) Ar->Cl 1 2 3 4 5 N 82.0 82.5 83.0 83.5 84.0 Accuracy (%) Office-Home (Cl->Ar) Cl->Ar [PITH_FULL_IMAGE:figures/full_fig_p016_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Sensitivity of parameters N. DIFO++ over DIFO. To further validate this obser￾vation, we present the 3D density chart results at the bottom of [PITH_FULL_IMAGE:figures/full_fig_p016_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Dynamics of gap-region reduction in epoch-wise view [PITH_FULL_IMAGE:figures/full_fig_p017_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Convergence behavior comparison on VisDA during the adaptation phase. DIFO++ cannot be deployed in black-box or cloud￾based settings where model access is restricted. Second, DIFO++ lacks an anti-forgetting mechanism, making it unsuitable for continuous learning scenarios. Third, the referenced entropy and gap region labeling depend on historical information accumulated during adapta￾tion, limiting its ap… view at source ↗
read the original abstract

Source-Free Domain Adaptation (SFDA) seeks to adapt a source model, which is pre-trained on a supervised source domain, for a target domain, with only access to unlabeled target training data. Relying on pseudo labeling and/or auxiliary supervision, conventional methods are inevitably error-prone. To mitigate this limitation, in this work we for the first time explore the potentials of off-the-shelf vision-language (ViL) multimodal models (e.g., CLIP) with rich whilst heterogeneous knowledge. We find that directly applying the ViL model to the target domain in a zero-shot fashion is unsatisfactory, as it is not specialized for this particular task but largely generic. To make it task-specific, we propose a novel DIFO++ approach. Specifically, DIFO++ alternates between two steps during adaptation: (i) Customizing the ViL model by maximizing the mutual information with the target model in a prompt learning manner, (ii) Distilling the knowledge of this customized ViL model to the target model, centering on gap region reduction. During progressive knowledge adaptation, we first identify and focus on the gap region, where enclosed features are entangled and class-ambiguous, as it often captures richer task-specific semantics. Reliable pseudo-labels are then generated by fusing predictions from the target and ViL models, supported by a memory mechanism. Finally, gap region reduction is guided by category attention and predictive consistency for semantic alignment, complemented by referenced entropy minimization to suppress uncertainty. Extensive experiments show that DIFO++ significantly outperforms the state-of-the-art alternatives. Our code and data are available at https://github.com/tntek/DIFO-Plus.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The manuscript introduces DIFO++ for source-free domain adaptation (SFDA). It alternates between (i) prompt-based customization of an off-the-shelf vision-language model (e.g., CLIP) by maximizing mutual information with the target model's predictions and (ii) distilling the customized ViL knowledge back to the target model via gap-region reduction, fused pseudo-labels, a memory bank, category attention, predictive consistency, and referenced entropy minimization. The central claim is that this progressive loop yields reliable task-specific knowledge transfer and significantly outperforms prior SFDA methods.

Significance. If the empirical gains hold, the work would be significant for SFDA by demonstrating how heterogeneous ViL priors can be specialized without source data or explicit supervision. Strengths include the explicit focus on gap regions for richer semantics, the memory-supported fusion mechanism, and the public release of code and data at https://github.com/tntek/DIFO-Plus, which aids reproducibility.

major comments (1)
  1. [Abstract and method description of the two-step alternation] The bootstrap assumption in the adaptation loop is load-bearing: the target model starts as a source-only network whose initial predictions on unlabeled target data are expected to be noisy. Maximizing mutual information against these predictions to customize the ViL model (step i) therefore risks supplying a biased supervision signal, after which distillation (step ii) may propagate the error. While the abstract and method description invoke progressive adaptation, gap-region focus, fused pseudo-labels, and memory to mitigate accumulation, no explicit analysis, ablation, or early-iteration diagnostics (e.g., accuracy curves or sensitivity to target-model initialization) are provided to substantiate that the loop remains stable.
minor comments (2)
  1. [Abstract] The abstract asserts 'significantly outperforms the state-of-the-art alternatives' without naming the datasets, reporting any numerical margins, or listing the baselines; this makes the strength of the central claim difficult to gauge from the summary alone.
  2. [Method section describing gap-region reduction] Notation for the gap region, category attention, and referenced entropy terms is introduced without an accompanying equation or diagram that would allow a reader to verify the exact objective being optimized in the distillation step.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback on our manuscript. We address the major comment below and will revise the paper to incorporate additional supporting analyses.

read point-by-point responses

  1. Referee: [Abstract and method description of the two-step alternation] The bootstrap assumption in the adaptation loop is load-bearing: the target model starts as a source-only network whose initial predictions on unlabeled target data are expected to be noisy. Maximizing mutual information against these predictions to customize the ViL model (step i) therefore risks supplying a biased supervision signal, after which distillation (step ii) may propagate the error. While the abstract and method description invoke progressive adaptation, gap-region focus, fused pseudo-labels, and memory to mitigate accumulation, no explicit analysis, ablation, or early-iteration diagnostics (e.g., accuracy curves or sensitivity to target-model initialization) are provided to substantiate that the loop remains stable.

    Authors: We appreciate the referee's observation that the bootstrap phase relies on initially noisy target predictions, which could in principle introduce bias into the mutual information maximization step and subsequent distillation. The method is designed to counteract this through its progressive alternation: the ViL customization step leverages the model's broad priors to regularize the target predictions, while the distillation step explicitly targets gap regions (where semantics are richer but ambiguous), employs fused pseudo-labels, a memory bank for temporal consistency, category attention, predictive consistency constraints, and referenced entropy minimization. These components are intended to enable gradual refinement rather than unchecked error propagation. That said, we agree that the manuscript would be strengthened by explicit empirical diagnostics of loop stability. In the revision we will add iteration-wise accuracy curves on the target domain, an ablation on the number of alternation cycles, and sensitivity experiments varying the source-only initialization to demonstrate convergence behavior and limited error accumulation. revision: yes

Circularity Check

0 steps flagged

No circularity: iterative empirical adaptation loop is self-contained

full rationale

The paper describes DIFO++ as an alternating iterative process of (i) prompt-based mutual information maximization to customize an off-the-shelf ViL model using the current target model and (ii) distillation back to the target model via gap-region reduction, pseudo-label fusion, and entropy minimization. No equations or algorithmic steps reduce any central claim to a fitted parameter defined in terms of itself, nor to a self-citation chain that bears the uniqueness or derivation load. The method relies on external pretrained ViL models and unlabeled target data as independent inputs; the bootstrap from initial noisy predictions is an empirical assumption, not a definitional circularity. The derivation chain therefore remains non-circular and externally grounded.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that vision-language models hold transferable heterogeneous knowledge that mutual information maximization can specialize without source supervision; no free parameters or invented entities are explicitly detailed in the abstract.

axioms (1)
  • domain assumption Off-the-shelf vision-language models like CLIP contain rich whilst heterogeneous knowledge that can be made task-specific for a target domain.
    Invoked to justify using ViL models as the starting point for customization in the absence of source data.

pith-pipeline@v0.9.0 · 5606 in / 1178 out tokens · 28497 ms · 2026-05-10T05:27:32.827042+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

82 extracted references · 5 canonical work pages

  1. [1]

    , Chen, S

    andonian2022robust APACrefauthors Andonian, A. , Chen, S. \ Hamid, R. APACrefauthors \ 2022 . Robust cross-modal representation learning with progressive self-distillation Robust cross-modal representation learning with progressive self-distillation . Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition Proceedings of IEEE/CVF con...

    2022

  2. [2]

    , Amirkhani, A

    banitalebi2023ebcdet APACrefauthors Banitalebi-Dehkordi, A. , Amirkhani, A. \ Mohammadinasab, A. APACrefauthors \ 2023 . EBCDet : Energy-Based Curriculum for Robust Domain Adaptive Object Detection EBCDet : Energy-based curriculum for robust domain adaptive object detection . IEEE Access 11 77810-77825

    2023

  3. [3]

    , Zhang, X

    Bi2022Entropyweighted APACrefauthors Bi, X. , Zhang, X. , Wang, S. \ Zhang, H. APACrefauthors \ 2022 . Entropy-weighted reconstruction adversary and curriculum pseudo labeling for domain adaptation in semantic segmentation Entropy-weighted reconstruction adversary and curriculum pseudo labeling for domain adaptation in semantic segmentation . Neurocomputi...

    2022

  4. [4]

    APACrefauthors \ 2006

    ent7bishop2006pattern APACrefauthors Bishop, C M. APACrefauthors \ 2006 . Pattern Recognition and Machine Learning Pattern recognition and machine learning . New York Springer

    2006

  5. [5]

    , Patwari, K

    chen2024empowering APACrefauthors Chen, D. , Patwari, K. , Lai, Z. , Cheung, S c. \ Chuah, C N. APACrefauthors \ 2024 . Empowering Source-Free Domain Adaptation with MLLM-driven Curriculum Learning Empowering source-free domain adaptation with mllm-driven curriculum learning . arXiv preprint arXiv:2405.18376

    work page arXiv 2024

  6. [6]

    chen2024sharegpt4v APACrefauthors Chen, L. , Li, J. , Dong, X. , Zhang, P. , He, C. , Wang, J. Lin, D. APACrefauthors \ 2024 . ShareGPT4v : Improving large multi-modal models with better captions ShareGPT4v : Improving large multi-modal models with better captions . Proceedings of the European Conference on Computer Vision Proceedings of the european conf...

    2024

  7. [7]

    , Zheng, Z

    chen2024transferring APACrefauthors Chen, M. , Zheng, Z. \ Yang, Y. APACrefauthors \ 2024 . Transferring to real-world layouts: A depth-aware framework for scene adaptation Transferring to real-world layouts: A depth-aware framework for scene adaptation . Proceedings of the ACM International Conference on Multimedia Proceedings of the acm international co...

    2024

  8. [8]

    , Zheng, Z

    chen2023pipa APACrefauthors Chen, M. , Zheng, Z. , Yang, Y. \ Chua, T S. APACrefauthors \ 2023 . Pipa: Pixel-and patch-wise self-supervised learning for domain adaptative semantic segmentation Pipa: Pixel-and patch-wise self-supervised learning for domain adaptative semantic segmentation . Proceedings of the ACM International Conference on Multimedia Proc...

    2023

  9. [9]

    , Lin, L

    chen2022self APACrefauthors Chen, W. , Lin, L. , Yang, S. , Xie, D. , Pu, S. \ Zhuang, Y. APACrefauthors \ 2022 . Self-supervised noisy label learning for source-free unsupervised domain adaptation Self-supervised noisy label learning for source-free unsupervised domain adaptation . IEEE/RSJ International Conference on Intelligent Robots and Systems IEEE/...

    2022

  10. [10]

    , Beaumont, R

    cherti2023reproducible APACrefauthors Cherti, M. , Beaumont, R. , Wightman, W. , Wortsman, M. , Ilharco, G. , Gordon, C. \ Jitsev, J. APACrefauthors \ 2023 . Reproducible scaling laws for contrastive language-image learning Reproducible scaling laws for contrastive language-image learning . Proceedings of the IEEE/CVF Conference on Computer Vision and Pat...

    2023

  11. [11]

    , Nam, G

    Cho_2023_ICCV APACrefauthors Cho, J. , Nam, G. , Kim, S. , Yang, H. \ Kwak, S. APACrefauthors \ 2023 . Promptstyler: Prompt-driven style generation for source-free domain generalization Promptstyler: Prompt-driven style generation for source-free domain generalization . Proceedings of the IEEE/CVF International Conference on Computer Vision Proceedings of...

    2023

  12. [12]

    , Jeong, M

    choi2019pseudo APACrefauthors Choi, J. , Jeong, M. , Kim, T. \ Kim, C. APACrefauthors \ 2019 . Pseudo-labeling curriculum for unsupervised domain adaptation Pseudo-labeling curriculum for unsupervised domain adaptation . arXiv preprint arXiv:1908.00262

    work page arXiv 2019

  13. [13]

    \ Thomas, J A

    cover2006elements APACrefauthors Cover, T M. \ Thomas, J A. APACrefauthors \ 2006 . Elements of information theory, 2nd ed Elements of information theory, 2nd ed . Hoboken, NJ, USA Wiley-Interscience

    2006

  14. [14]

    ding2022source APACrefauthors Ding, N. , Xu, Y. , Tang, Y. , Xu, C. , Wang, Y. \ Tao, D. APACrefauthors \ 2022 . Source-free domain adaptation via distribution estimation Source-free domain adaptation via distribution estimation . Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Proceedings of the IEEE/CVF conference on co...

    2022

  15. [15]

    , Yang, H

    du2021ps APACrefauthors Du, Y. , Yang, H. , Chen, M. , Jiang, J. , Luo, H. \ Wang, C. APACrefauthors \ 2021 . Generation, augmentation, and alignment: A pseudo-source domain based method for source-free domain adaptation Generation, augmentation, and alignment: A pseudo-source domain based method for source-free domain adaptation . arXiv:2109.04015

    work page arXiv 2021

  16. [16]

    \ Copas, J

    eguchi2006interpreting APACrefauthors Eguchi, S. \ Copas, J. APACrefauthors \ 2006 . Interpreting kullback--leibler divergence with the neyman--pearson lemma Interpreting kullback--leibler divergence with the neyman--pearson lemma . Journal of Multivariate Analysis 97 9 2034--2040

    2006

  17. [17]

    \ Ghahramani, Z

    gal2016dropout APACrefauthors Gal, Y. \ Ghahramani, Z. APACrefauthors \ 2016 . Dropout as a bayesian approximation: Representing model uncertainty in deep learning Dropout as a bayesian approximation: Representing model uncertainty in deep learning . international conference on machine learning international conference on machine learning \ ( \ 1050--1059)

    2016

  18. [18]

    \ Lempitsky, V

    ganin2015unsupervised APACrefauthors Ganin, Y. \ Lempitsky, V. APACrefauthors \ 2015 . Unsupervised domain adaptation by backpropagation Unsupervised domain adaptation by backpropagation . Proceedings of the international conference on machine learning Proceedings of the international conference on machine learning \ ( \ 1180--1189)

    2015

  19. [19]

    , Tassi, C R N

    gawlikowski2023survey APACrefauthors Gawlikowski, J. , Tassi, C R N. , Ali, M. , Lee, J. , Humt, M. , Feng, J. others APACrefauthors \ 2023 . A survey of uncertainty in deep neural networks A survey of uncertainty in deep neural networks . Artificial intelligence review 56 Suppl 1 1513--1589

    2023

  20. [20]

    , Huang, R

    ge2025pluda APACrefauthors Ge, C. , Huang, R. , Xie, M. , Lai, Z. , Song, S. , Li, S. \ Huang, G. APACrefauthors \ 2025 . Domain Adaptation via Prompt Learning Domain adaptation via prompt learning . IEEE Transactions on Neural Networks and Learning Systems 36 1 1160-1170 . APACrefDOI doi:10.1109/TNNLS.2023.3327962 APACrefDOI

    work page doi:10.1109/tnnls.2023.3327962 2025

  21. [21]

    , Guan, D

    huang2021model APACrefauthors Huang, J. , Guan, D. , Xiao, A. \ Lu, S. APACrefauthors \ 2021 . Model adaptation: Historical contrastive learning for unsupervised domain adaptation without source data Model adaptation: Historical contrastive learning for unsupervised domain adaptation without source data . Advances in neural information processing systems ...

    2021

  22. [22]

    , Lee, J

    hwang2024sfda APACrefauthors Hwang, U. , Lee, J. , Shin, J. \ Yoon, S. APACrefauthors \ 2024 . SF ( DA ) 2 : Source-free Domain Adaptation Through the Lens of Data Augmentation SF ( DA ) 2 : Source-free domain adaptation through the lens of data augmentation . International Conference on Learning Representations. International conference on learning repre...

    2024

  23. [23]

    , Henriques, J F

    ji2019invariant APACrefauthors Ji, X. , Henriques, J F. \ Vedaldi, A. APACrefauthors \ 2019 . Invariant information clustering for unsupervised image classification and segmentation Invariant information clustering for unsupervised image classification and segmentation . Proceedings of the IEEE/CVF International Conference on Computer Vision Proceedings o...

    2019

  24. [24]

    jiang2025hg APACrefauthors Jiang, J. , Lv, Q. , Li, Y. , Du, Y. , Dong, J. , Chen, S. \ Yu, H. APACrefauthors \ 2025 . HG-SFDA: HyperGraph Learning Meets Source-Free Unsupervised Domain Adaptation Hg-sfda: Hypergraph learning meets source-free unsupervised domain adaptation . IEEE Transactions on Image Processing 34 7542--7557

    2025

  25. [25]

    , Jiang, L

    kang2019contrastive APACrefauthors Kang, G. , Jiang, L. , Yang, Y. \ Hauptmann, A G. APACrefauthors \ 2019 . Contrastive adaptation network for unsupervised domain adaptation Contrastive adaptation network for unsupervised domain adaptation . Proceedings of the IEEE/CVF conference on computer vision and pattern recognition Proceedings of the IEEE/CVF conf...

    2019

  26. [26]

    , Lal, R

    mtdaKumar2023WACV APACrefauthors Kumar, V. , Lal, R. , Patil, H. \ Chakraborty, A. APACrefauthors \ 2023 January . CoNMix for Source-Free Single and Multi-Target Domain Adaptation Conmix for source-free single and multi-target domain adaptation . Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) Proceedings of the iee...

    2023

  27. [27]

    , Kulkarni, A R

    kundu2022balancing APACrefauthors Kundu, J N. , Kulkarni, A R. , Bhambri, S. , Mehta, D. , Kulkarni, S A. , Jampani, V. \ Radhakrishnan, V B. APACrefauthors \ 2022 . Balancing discriminability and transferability for source-free domain adaptation Balancing discriminability and transferability for source-free domain adaptation . Proceedings of the internat...

    2022

  28. [28]

    , Subramanian, V K

    kurmi2021domain APACrefauthors Kurmi, V K. , Subramanian, V K. \ Namboodiri, V P. APACrefauthors \ 2021 . Domain impression: A source data free domain adaptation method Domain impression: A source data free domain adaptation method . Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision Proceedings of the IEEE/CVF winter confere...

    2021

  29. [29]

    , Zhou, Y

    h0lai2023memory APACrefauthors Lai, Y. , Zhou, Y. , Liu, X. \ Zhou, T. APACrefauthors \ 2024 . Memory-assisted sub-prototype mining for universal domain adaptation Memory-assisted sub-prototype mining for universal domain adaptation . International Conference on Learning Representations. International conference on learning representations

    2024

  30. [30]

    , Jiang, X

    lao2021hypothesis APACrefauthors Lao, Q. , Jiang, X. \ Havaei, M. APACrefauthors \ 2021 . Hypothesis disparity regularized mutual information maximization Hypothesis disparity regularized mutual information maximization . Proceedings of the AAAI conference on artificial intelligence Proceedings of the AAAI conference on artificial intelligence \ ( 35, \ 8...

    2021

  31. [31]

    , Jung, D

    lee2022confidence APACrefauthors Lee, J. , Jung, D. , Yim, J. \ Yoon, S. APACrefauthors \ 2022 . Confidence score for source-free unsupervised domain adaptation Confidence score for source-free unsupervised domain adaptation . Proceedings of the International Conference on Machine Learning Proceedings of the international conference on machine learning \ ...

    2022

  32. [32]

    , Park, J H

    ent2lee2025duet APACrefauthors Lee, J Y. , Park, J H. , Lee, G. , Kim, B. , Cha, M H. , Nam, H. Cho, S I. APACrefauthors \ 2024 . DUET: Dual-Perspective Pseudo Labeling and Uncertainty-aware Exploration & Exploitation Training for Source-Free Domain Adaptation Duet: Dual-perspective pseudo labeling and uncertainty-aware exploration & exploitation training...

    2024

  33. [33]

    li2025enhancing APACrefauthors Li, H. \ Li, B. APACrefauthors \ 2025 . Enhancing vision-language compositional understanding with multimodal synthetic data Enhancing vision-language compositional understanding with multimodal synthetic data . Proceedings of the Computer Vision and Pattern Recognition Conference Proceedings of the computer vision and patte...

    2025

  34. [34]

    li2021divergence APACrefauthors Li, J. , Du, Z. , Zhu, L. , Ding, Z. , Lu, K. \ Shen, H T. APACrefauthors \ 2021 . Divergence-agnostic unsupervised domain adaptation by adversarial attacks Divergence-agnostic unsupervised domain adaptation by adversarial attacks . IEEE Transactions on Pattern Analysis and Machine Intelligence 44 11 8196--8211

    2021

  35. [35]

    li2025unified APACrefauthors Li, X. , Li, J. , Du, Z. , Zhu, L. \ Shen, H T. APACrefauthors \ 2025 . Unified Modality Separation: A Vision-Language Framework for Unsupervised Domain Adaptation Unified modality separation: A vision-language framework for unsupervised domain adaptation . IEEE Transactions on Pattern Analysis and Machine Intelligence 47 11 1...

    2025

  36. [36]

    liang2023open APACrefauthors Liang, F. , Wu, B. , Dai, X. , Li, K. , Zhao, Y. , Zhang, H. Marculescu, D. APACrefauthors \ 2023 . Open-vocabulary semantic segmentation with mask-adapted clip Open-vocabulary semantic segmentation with mask-adapted clip . Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Proceedings of the IEE...

    2023

  37. [37]

    2019Distant APACrefauthors Liang, J. , He, R. , Sun, Z. \ Tan, T. APACrefauthors \ 2019 . Distant supervised centroid shift: A simple and efficient approach to visual domain adaptation Distant supervised centroid shift: A simple and efficient approach to visual domain adaptation . Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recog...

    2019

  38. [38]

    liang2020we APACrefauthors Liang, J. , Hu, D. \ Feng, J. APACrefauthors \ 2020 . Do we really need to access the source data? source hypothesis transfer for unsupervised domain adaptation Do we really need to access the source data? source hypothesis transfer for unsupervised domain adaptation . Proceedings of the International Conference on Machine Learn...

    2020

  39. [39]

    , Del Bue, A

    Litrico_2023_CVPR APACrefauthors Litrico, M. , Del Bue, A. \ Morerio, P. APACrefauthors \ 2023 . Guiding pseudo-labels with uncertainty estimation for source-free unsupervised domain adaptation Guiding pseudo-labels with uncertainty estimation for source-free unsupervised domain adaptation . Proceedings of the IEEE/CVF Conference on Computer Vision and Pa...

    2023

  40. [40]

    , Wang, X

    liu2020energy APACrefauthors Liu, W. , Wang, X. , Owens, J. \ Li, Y. APACrefauthors \ 2020 . Energy-based out-of-distribution detection Energy-based out-of-distribution detection . Advances in Neural Information Processing Systems Advances in neural information processing systems \ ( 33, \ 21464--21475)

    2020

  41. [41]

    , Cao, Z

    long2018conditional APACrefauthors Long, M. , Cao, Z. , Wang, J. \ Jordan, M I. APACrefauthors \ 2018 . Conditional adversarial domain adaptation Conditional adversarial domain adaptation . Advances in neural information processing systems Advances in neural information processing systems \ ( 31, \ 1647--1657)

    2018

  42. [42]

    , Izmailov, P

    maddox2019simple APACrefauthors Maddox, W J. , Izmailov, P. , Garipov, T. , Vetrov, D P. \ Wilson, A G. APACrefauthors \ 2019 . A simple baseline for bayesian uncertainty in deep learning A simple baseline for bayesian uncertainty in deep learning . Advances in neural information processing systems 32

    2019

  43. [43]

    , Bai, Q

    peng2019moment APACrefauthors Peng, X. , Bai, Q. , Xia, X. , Huang, Z. , Saenko, K. \ Wang, B. APACrefauthors \ 2019 . Moment matching for multi-source domain adaptation Moment matching for multi-source domain adaptation . Proceedings of the IEEE/CVF International Conference on Computer Vision Proceedings of the IEEE/CVF international conference on comput...

    2019

  44. [44]

    VisDA: The Visual Domain Adaptation Challenge

    peng2017visda APACrefauthors Peng, X. , Usman, B. , Kaushik, N. , Hoffman, J. , Wang, D. \ Saenko, K. APACrefauthors \ 2017 . Visda: The visual domain adaptation challenge Visda: The visual domain adaptation challenge . arXiv:1710.06924

    work page Pith review arXiv 2017

  45. [45]

    , Kim, J W

    radford2021learning APACrefauthors Radford, A. , Kim, J W. , Hallacy, C. , Ramesh, A. , Goh, G. , Agarwal, S. others APACrefauthors \ 2021 . Learning transferable visual models from natural language supervision Learning transferable visual models from natural language supervision . International conference on machine learning International conference on m...

    2021

  46. [46]

    , Thakur, R S

    rai2025label APACrefauthors Rai, S. , Thakur, R S. , Jangid, K. \ Kurmi, V K. APACrefauthors \ 2025 . Label Calibration in Source Free Domain Adaptation Label calibration in source free domain adaptation . IEEE/CVF Winter Conference on Applications of Computer Vision IEEE/CVF winter conference on applications of computer vision \ ( \ 6446--6455)

    2025

  47. [47]

    , Trapp, M

    roy2022uncertainty APACrefauthors Roy, S. , Trapp, M. , Pilzer, A. , Kannala, J. , Sebe, N. , Ricci, E. \ Solin, A. APACrefauthors \ 2022 . Uncertainty-guided source-free domain adaptation Uncertainty-guided source-free domain adaptation . Proceedings of the European conference on computer vision Proceedings of the european conference on computer vision \...

    2022

  48. [48]

    , Kulis, B

    saenko2010adapting APACrefauthors Saenko, K. , Kulis, B. , Fritz, M. \ Darrell, T. APACrefauthors \ 2010 . Adapting visual category models to new domains Adapting visual category models to new domains . Proceedings of the European Conference on Computer Vision Proceedings of the european conference on computer vision \ ( \ 213--226)

    2010

  49. [49]

    ent5safaei2025certainty APACrefauthors Safaei, B. , VS, V. \ Patel, V M. APACrefauthors \ 2025 . Certainty and Uncertainty Guided Active Domain Adaptation Certainty and uncertainty guided active domain adaptation . IEEE International Conference on Image Processing. Ieee international conference on image processing

    2025

  50. [50]

    , Kim, D

    saito2019semi APACrefauthors Saito, K. , Kim, D. , Sclaroff, S. , Darrell, T. \ Saenko, K. APACrefauthors \ 2019 . Semi-supervised domain adaptation via minimax entropy Semi-supervised domain adaptation via minimax entropy . Proceedings of the IEEE/CVF international conference on computer vision Proceedings of the IEEE/CVF international conference on comp...

    2019

  51. [51]

    , Woo, S

    r4shin2020two APACrefauthors Shin, I. , Woo, S. , Pan, F. \ Kweon, I S. APACrefauthors \ 2020 . Two-phase pseudo label densification for self-training based domain adaptation Two-phase pseudo label densification for self-training based domain adaptation . European Conference on Computer Vision European conference on computer vision \ ( \ 532--548)

    2020

  52. [52]

    , Fan, J

    f0song2024multi APACrefauthors Song, Y. , Fan, J. , Liu, D. \ Cai, W. APACrefauthors \ 2024 . Multi-source-free domain adaptation via uncertainty-aware adaptive distillation Multi-source-free domain adaptation via uncertainty-aware adaptive distillation . 2024 IEEE International Symposium on Biomedical Imaging (ISBI) 2024 ieee international symposium on b...

    2024

  53. [53]

    , Chang, A

    tang2023source APACrefauthors Tang, S. , Chang, A. , Zhang, F. , Zhu, X. , Ye, M. \ Zhang, C. APACrefauthors \ 2024 . Source-Free Domain Adaptation via Target Prediction Distribution Searching Source-free domain adaptation via target prediction distribution searching . International Journal of Computer Vision 132 654–672

    2024

  54. [54]

    , Shi, Y

    tang2021model APACrefauthors Tang, S. , Shi, Y. , Ma, Z. , Li, J. , Lyu, J. , Li, Q. \ Zhang, J. APACrefauthors \ 2021 . Model adaptation through hypothesis transfer with gradual knowledge distillation Model adaptation through hypothesis transfer with gradual knowledge distillation . IEEE/RSJ International Conference on Intelligent Robots and Systems (IRO...

    2021

  55. [55]

    tang2025proxy APACrefauthors Tang, S. , Su, W. , Gan, Y. , Ye, M. , Zhang, J. \ Zhu, X. APACrefauthors \ 2025 . Proxy Denoising for Source-Free Domain Adaptation Proxy denoising for source-free domain adaptation . International Conference on Learning Representations. International conference on learning representations. APACrefURL https://openreview.net/f...

    2025

  56. [56]

    tang2024sourcefree APACrefauthors Tang, S. , Su, W. , Ye, M. \ Zhu, X. APACrefauthors \ 2024 . Source-free domain adaptation with frozen multimodal foundation model Source-free domain adaptation with frozen multimodal foundation model . Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Proceedings of the IEEE/CVF conference...

    2024

  57. [57]

    tang2019adaptive APACrefauthors Tang, S. , Ye, M. , Xu, P. \ Li, X. APACrefauthors \ 2019 . Adaptive pedestrian detection by predicting classifier Adaptive pedestrian detection by predicting classifier . Neural Computing and Applications 31 1189--1200

    2019

  58. [58]

    , Zou, Y

    tang2022sclm APACrefauthors Tang , S. , Zou, Y. , Song, Z. , Lyu, J. , Chen, L. , Ye, M. Zhang, J. APACrefauthors \ 2022 . Semantic consistency learning on manifold for source data-free unsupervised domain adaptation Semantic consistency learning on manifold for source data-free unsupervised domain adaptation . Neural Networks 152 467-478

    2022

  59. [59]

    , Fan, X

    tanwisuth2021pct APACrefauthors Tanwisuth, K. , Fan, X. , Zheng, H. , Zhang, S. , Zhang, H. , Chen, B. \ Zhou, M. APACrefauthors \ 2021 . A prototype-oriented framework for unsupervised domain adaptation A prototype-oriented framework for unsupervised domain adaptation . Advances in Neural Information Processing Systems Advances in neural information proc...

    2021

  60. [60]

    Attracting and Dispersing

    tarashima2025vilaad APACrefauthors Tarashima, S. , Shu, X. \ Tagawa, N. APACrefauthors \ 2025 . ViLAaD : Enhancing" Attracting and Dispersing" Source-Free Domain Adaptation with Vision-and-Language Model ViLAaD : Enhancing" attracting and dispersing" source-free domain adaptation with vision-and-language model . Neural Information Processing Systems, Mult...

    2025

  61. [61]

    , Zhang, J

    tian2021vdm APACrefauthors Tian, J. , Zhang, J. , Li, W. \ Xu, D. APACrefauthors \ 2021 1 . VDM-DA: Virtual domain modeling for source data-free domain adaptation Vdm-da: Virtual domain modeling for source data-free domain adaptation . IEEE Transactions on Circuits and Systems for Video Technology 32 6 3749--3760

    2021

  62. [62]

    , Zhang, J

    tian2022vdm APACrefauthors Tian, J. , Zhang, J. , Li, W. \ Xu, D. APACrefauthors \ 2021 2 . VDM-DA : Virtual domain modeling for source data-free domain adaptation VDM-DA : Virtual domain modeling for source data-free domain adaptation . IEEE Transactions on Circuits and Systems for Video Technology 32 6 3749--3760

    2021

  63. [63]

    , Eusebio, J

    venkateswara2017deep APACrefauthors Venkateswara, H. , Eusebio, J. , Chakraborty, S. \ Panchanathan, S. APACrefauthors \ 2017 . Deep hashing network for unsupervised domain adaptation Deep hashing network for unsupervised domain adaptation . Proceedings of the IEEE conference on computer vision and pattern recognition Proceedings of the ieee conference on...

    2017

  64. [64]

    , Han, Z

    wang2022exploring APACrefauthors Wang, F. , Han, Z. , Gong, Y. \ Yin, Y. APACrefauthors \ 2022 . Exploring domain-invariant parameters for source free domain adaptation Exploring domain-invariant parameters for source free domain adaptation . Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Proceedings of the IEEE/CVF conf...

    2022

  65. [65]

    , Bae, W

    wang2025silan APACrefauthors Wang, J. , Bae, W. , Chen, J. , Zhang, K. , Sigal, L. \ de Silva, C W. APACrefauthors \ 2025 . What Has Been Overlooked in Contrastive Source-Free Domain Adaptation: Leveraging Source-Informed Latent Augmentation within Neighborhood Context What has been overlooked in contrastive source-free domain adaptation: Leveraging sourc...

    2025

  66. [66]

    , Fink, O

    continualwang2022continual APACrefauthors Wang, Q. , Fink, O. , Van Gool, L. \ Dai, D. APACrefauthors \ 2022 . Continual test-time domain adaptation Continual test-time domain adaptation . Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Proceedings of the ieee/cvf conference on computer vision and pattern recognition \ ( ...

    2022

  67. [67]

    , Zhang, Z

    mai_wang2024consistency APACrefauthors Wang, Y. , Zhang, Z. , Jin, X. \ Zeng, W. APACrefauthors \ 2024 . Consistency Prior Matters: Biomedical-Prompting Dual Augmentation for Domain Adaptive Medical Image Segmentation Consistency prior matters: Biomedical-prompting dual augmentation for domain adaptive medical image segmentation . 2024 IEEE International ...

    2024

  68. [68]

    , Zhao, H

    xia2021adaptive APACrefauthors Xia, H. , Zhao, H. \ Ding, Z. APACrefauthors \ 2021 . Adaptive adversarial network for source-free domain adaptation Adaptive adversarial network for source-free domain adaptation . Proceedings of the IEEE/CVF International Conference on Computer Vision Proceedings of the IEEE/CVF international conference on computer vision ...

    2021

  69. [69]

    , Guo, H

    ent1xu2025revisiting APACrefauthors Xu, G. , Guo, H. , Yi, L. , Ling, C. , Wang, B. \ Yi, G. APACrefauthors \ 2024 . Revisiting Source-Free Domain Adaptation: a New Perspective via Uncertainty Control Revisiting source-free domain adaptation: a new perspective via uncertainty control . International Conference on Learning Representations. International co...

    2024

  70. [70]

    , Guo, H

    xu2025revisiting APACrefauthors Xu, G. , Guo, H. , Yi, L. , Ling, C. , Wang, B. \ Yi, G. APACrefauthors \ 2025 . Revisiting Source-Free Domain Adaptation: a New Perspective via Uncertainty Control Revisiting source-free domain adaptation: a new perspective via uncertainty control . International Conference on Learning Representations. International confer...

    2025

  71. [71]

    xu2019larger APACrefauthors Xu, R. , Li, G. , Yang, J. \ Lin, L. APACrefauthors \ 2019 . Larger norm more transferable: An adaptive feature norm approach for unsupervised domain adaptation Larger norm more transferable: An adaptive feature norm approach for unsupervised domain adaptation . Proceedings of the IEEE/CVF International Conference on Computer V...

    2019

  72. [72]

    , Peng, X

    r2yang2022divide APACrefauthors Yang, J. , Peng, X. , Wang, K. , Zhu, Z. , Feng, J. , Xie, L. \ You, Y. APACrefauthors \ 2023 . Divide to adapt: Mitigating confirmation bias for domain adaptation of black-box predictors Divide to adapt: Mitigating confirmation bias for domain adaptation of black-box predictors . International Conference on Learning Repres...

    2023

  73. [73]

    , Jui, S

    yang2022attracting APACrefauthors Yang, S. , Jui, S. , Van De Weijer, J. \ . APACrefauthors \ 2022 . Attracting and dispersing: A simple approach for source-free domain adaptation Attracting and dispersing: A simple approach for source-free domain adaptation . Advances in Neural Information Processing Systems Advances in neural information processing syst...

    2022

  74. [74]

    , Wang, Y

    i0yang2021exploiting APACrefauthors Yang, S. , Wang, Y. , Joost, V D W. , Luis, H. \ Jui, S. APACrefauthors \ 2021 . Exploiting the intrinsic neighborhood structure for source-free domain adaptation Exploiting the intrinsic neighborhood structure for source-free domain adaptation . Advances in Neural Information Processing Systems Advances in neural infor...

    2021

  75. [75]

    yi2023source APACrefauthors Yi, L. , Xu, G. , Xu, P. , Li, J. , Pu, R. , Ling, C. Wang, B. APACrefauthors \ 2023 . When Source-Free Domain Adaptation Meets Learning with Noisy Labels When source-free domain adaptation meets learning with noisy labels . International Conference on Learning Representations. International conference on learning representatio...

    2023

  76. [76]

    , Yao, Z

    g0yu2025smdanet APACrefauthors Yu, Z. , Yao, Z. , Wang, W. , Jiang, Q. \ Cao, Z. APACrefauthors \ 2025 . SmdaNet : A hierarchical hard sample mining and domain adaptation neural network for fault diagnosis in industrial process SmdaNet : A hierarchical hard sample mining and domain adaptation neural network for fault diagnosis in industrial process . Chin...

    2025

  77. [77]

    , Shen, L

    zhang2025source APACrefauthors Zhang, W. , Shen, L. \ Foo, C S. APACrefauthors \ 2025 . Source-free domain adaptation guided by vision and vision-language pre-training Source-free domain adaptation guided by vision and vision-language pre-training . International Journal of Computer Vision 133 2 844--866

    2025

  78. [78]

    , Wang, Z

    zhang2023class APACrefauthors Zhang, Y. , Wang, Z. \ He, W. APACrefauthors \ 2023 . Class relationship embedded learning for source-free unsupervised domain adaptation Class relationship embedded learning for source-free unsupervised domain adaptation . Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Proceedings of the IE...

    2023

  79. [79]

    , Yang, J

    zhou2022learning APACrefauthors Zhou, K. , Yang, J. , Loy, C C. \ Liu, Z. APACrefauthors \ 2022 . Learning to prompt for vision-language models Learning to prompt for vision-language models . International Journal of Computer Vision 130 9 2337--2348

    2022

  80. [80]

    ZHOU2024109974 APACrefauthors Zhou, L. , Li, N. , Ye, M. , Zhu, X. \ Tang, S. APACrefauthors \ 2024 . Source-free domain adaptation with Class Prototype Discovery Source-free domain adaptation with class prototype discovery . Pattern Recognition 145 109974

    2024

Showing first 80 references.