Pith. sign in

REVIEW 5 major objections 5 minor 52 references

TransAdapter: Vision Transformer for Feature-Centric Unsupervised Domain Adaptation

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that TransAdapter, a Swin Transformer modified with graph domain discrimination, adaptive double attention, and cross-feature transform, reaches state-of-the-art accuracy on four unsupervised domain adaptation benchmarks.

desk verdict A plausible UDA architecture whose central CFT equations don't typecheck; the benchmark claims are untraceable until the implementation is clarified. read the letter →

arxiv 2412.04073 v1 pith:WFYQRMC4 submitted 2024-12-05 cs.CV

A. Enes Doruk , Erhan Oztop , Hasan F. Ates This is my paper · ORCID
classification cs.CV
keywords UnsupervisedDomainAdaptationVisionTransformerSwinGraphDiscriminatorAdaptiveDoubleAttentionCrossFeatureTransformPseudo-labelingbenchmarks
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that replacing the fixed attention of a Swin Transformer with three domain-alignment modules is enough to make a single UDA method dominate four standard benchmarks without task-specific alignment heads. It reports average accuracies of 95.5% on Office-31, 89.4% on Office-Home, 91.2% on VisDA-2017, and 53.7% on DomainNet, all above the previous transformer-based UDA methods it compares against. If the claims hold, a generic feature-centric transformer recipe can replace bespoke adaptation modules and keep improving when backbones grow. The paper also argues each module contributes independently, with the cross-feature transform giving the largest single jump on VisDA-2017 in the ablation.

What carries the argument

The load-bearing identity is the modified Swin block: instead of window attention alone, the ADA module computes window and shifted-window attention maps, reweights them with entropy $H(F_{graph})=-\sum_i F_{graph}\log F_{graph}$ from the graph discriminator, concatenates and softmaxes them, and applies them to concatenated value vectors. The Graph Domain Discriminator supplies the entropy signal and an adversarial gradient through graph convolutional layers acting on an adjacency $P(x_i)\cdot P(x_j)/(|P(x_i)||P(x_j)|)$. The Cross Feature Transform computes $F_{s2t}=\mathrm{Softmax}(f(X_s)^\top g(X_t))$ and $F_{t2s}=\mathrm{Softmax}(g(X_t)^\top f(X_s))$, blends them with a sigmoid-gated parameter $\gamma$, and, as written, adds $\mathrm{Attention}_{gating}\times\|F_{s2t}-F_{t2s}\|_2^2$ to $X_t$. That last combination is the step whose dimensions do not match as printed, making the exact mechanism contestable at the implementation level.

What would settle it

Inspect the released forward pass of the Cross Feature Transform module in Section 3.3: if the tensor shapes in Eq. (8) are as described, the operation will not run, and the actual implementation must be recovered from code. Re-running the ablation with a corrected pairwise-distance term and comparing the VisDA-2017 and DomainNet gains would settle whether the claimed mechanism causes the improvement.

Watch

Extended reading notes

Core claim

On its own terms, the central discovery is that a Swin Transformer can be adapted to an unlabeled target domain by (1) a Graph Domain Discriminator that builds a cosine-similarity adjacency matrix over source and target key features and uses graph convolutions with gradient reversal to align relational structure, (2) an Adaptive Double Attention module that fuses window and shifted-window attention and reweights attention scores by an entropy map derived from the discriminator, and (3) a Cross Feature Transform that mixes source and target features with gated bidirectional cross-attention plus a pairwise-distance term. Trained with pseudo-label-guided CutMix and MixUp on the source side, the resulting TransAdapter-B outperforms the prior state of the art on Office-31, Office-Home, VisDA-2017, and DomainNet.

Load-bearing premise

The load-bearing premise is that the Cross Feature Transform can be built as the paper's Eq. (8) describes; that formula adds a square table of attention weights to a rectangle of image features, so the written operation cannot run as stated. The reported gains from this module therefore rest on an unstated implementation detail.

Editorial extensions

If this is right

  • If the reported numbers reproduce, a feature-centric transformer with no task-specific alignment head would become the reference point for UDA on Office-Home and VisDA-2017.
  • The graph discriminator's entropy map becomes a reusable alignment signal that any transformer attention could be reweighted by, not just Swin-style windows.
  • Pseudo-label-guided CutMix and MixUp applied only to source data would be sufficient to add one to two points on top of alignment losses, reducing the need for target-side augmentation.
  • Because each module shows an additive gain in the ablation, the same three-module recipe should transfer to other hierarchical vision transformers.
  • If the CFT dimension issue is resolved in the released code, the pairwise-distance term is the first place to look for the claimed +4.8% VisDA-2017 and +2.4% DomainNet gains.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • My read is that the as-printed Eq. (8) cannot run as stated because an $N\times N$ attention/distance matrix is added to an $N\times D$ feature tensor; the actual operation in the released implementation is the decisive unknown for attributing the CFT ablation gains.
  • A natural test the paper does not run is ablating the pairwise-distance term alone while keeping the gated cross-attention; this would show whether the distance signal or the gating carries the improvement.
  • The entropy-reweighting mechanism could be applied outside UDA, such as downweighting high-entropy attention in out-of-distribution detection or spurious-correlation rejection, since it is designed to suppress domain-specific patterns.
  • Because the pseudo-labels come from a Swin-Base teacher, the method's ceiling may depend on teacher quality; feeding TransAdapter's own confident predictions back into the confidence threshold would test a self-distillation extension.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper proposes TransAdapter, a Swin Transformer-based unsupervised domain adaptation method with three modules: a Graph Domain Discriminator (GDD), an Adaptive Double Attention (ADA) module, and a Cross-Feature Transform (CFT), together with CutMix/MixUp augmentation guided by pseudo-labels. The authors report state-of-the-art results on Office-31 (95.5%), Office-Home (89.4%), VisDA-2017 (91.2%), and DomainNet (53.7%), and provide ablations and t-SNE visualizations. The central claim is that inserting these modules into Swin Transformer blocks yields consistent and large improvements over strong baselines.

Significance. If fully realizable and reproducible, the proposed combination of graph-based discrimination, dual attention, and cross-feature transformation would be a useful contribution to transformer-based UDA, and the benchmark coverage is broad. The paper compares with many recent methods and reports per-category results. However, the significance is not established from the manuscript as written: the CFT equations are dimensionally inconsistent, the entropy reweighting in ADA appears to have the wrong sign relative to its stated purpose, the GDD is described simultaneously at per-sample and per-token levels without a bridge, and the objective in Eq. (10) is not a well-defined loss without an additional discriminator head. Because these issues concern the modules that carry the reported gains, the claimed state-of-the-art performance cannot be verified from the text alone. The paper does provide useful strengths: clear architectural figures, a broad experimental comparison, explicit hyperparameter choices, and an intention to release code.

major comments (5)
  1. [Section 3.3, Eq. (6)-(8)] The CFT module as written is not realizable. If Xs and Xt are N×D feature matrices, then f(Xs)^T g(Xt) is D×D, so Fs2t and Ft2s are D×D matrices; Attngating in Eq. (7) is then also D×D. The pairwise-distance term ||Fs2t − Ft2s||^2_2 is D×D, so Attngating × ||Fs2t − Ft2s||^2_2 remains D×D, while Xt is N×D; the addition in Eq. (8) is therefore dimensionally invalid. If a different layout is intended, the shapes must be defined precisely, including how the module operates per token or per window. As written, the ablation gains attributed to CFT in Table 3 (+4.8 on VisDA-2017, +2.4 on DomainNet) cannot be traced to the described operation.
  2. [Section 3.1, Eq. (1)-(2)] The entropy-guided reweighting is internally inconsistent. Eq. (1) defines H(Fgraph) as a scalar entropy summed over feature entries for each sample, but Eq. (2) multiplies the attention matrix A elementwise by H(Fgraph), which requires H to have the same shape as the attention map. No definition is given for how the per-sample scalar becomes an attention-shaped matrix. Moreover, the text states that higher entropy signals domain-specific noise that should be suppressed, yet multiplying attention by H would emphasize high-entropy entries. The intended weighting is presumably a decreasing function of entropy (for example 1−H or exp(−H)), or H must be normalized and inverted; the paper needs to correct this and define the shape of H.
  3. [Section 3.2] The GDD is described at two incompatible levels. Eq. (5) builds an adjacency matrix over samples and the graph convolutional layers aggregate information across samples, so the GDD output is necessarily batch-level. But Figs. 2–3 and Eq. (2) use Fs_graph and Ft_graph as per-token, per-window entropy inputs inside each Swin block. The paper does not explain how a batch-level, sample-mixing graph output is reduced back to per-window attention maps of the same shape as A and Ashift. This conflation between per-sample and per-token organization makes the GDD–ADA interaction unimplementable as written.
  4. [Section 4.3, Eq. (10)-(11)] The adaptation loss is not well defined. Eq. (10) computes cross-entropy (local) and focal loss (global) between a feature output F_ADV and constant labels ŷsrc=1, ŷtgt=0, but F_ADV is described as the output of the second or last transformer block, not as the output of a domain-discriminator head. Unless F_ADV is first passed through a two-class classifier, cross-entropy on raw transformer features with scalar labels is not a meaningful loss. The Gradient Reversal Layer mentioned in Sec. 3.2 is also not connected to this objective. The total loss in Eq. (11) is therefore not reproducible from the text.
  5. [Section 4.4 / Tables 1 and 3] The experimental evidence is reported without variance: no standard deviations, number of seeds, or repeated runs are given for any table, including close margins such as Office-Home 89.4 vs. 89.0 (Table 1) and Office-31 95.5 vs. 95.3 (Table 5). In addition, the final TransAdapter-B Office-Home accuracy in the ablation Table 3 is 87.5, while Table 1 reports 89.4 for the same configuration. The pseudo-label confidence threshold in Sec. 3.4 is never specified, and no sensitivity analysis is provided. These gaps prevent verification of the reported improvements and of the claimed state-of-the-art status.
minor comments (5)
  1. [Section 3.2, Eq. (5)] The projection P in Eq. (5) is not defined; the paper should state its form and how it relates to the source and target key features Ks and Kt shown in Fig. 3.
  2. [Section 3.1 / Figure 1] The symbol H is used both for the entropy matrix in Eq. (1)–(2) and for the feature-map height in Fig. 1, which is confusing.
  3. [Table 2] The DomainNet table is poorly formatted in the text, with method names, column headers, and row labels interleaved; readers cannot easily verify the per-domain numbers and averages.
  4. [Section 4.2] The statement 'our code available at enesdoruk/TransAdapter' is not a URL; a working repository link should be provided.
  5. [Section 5] The conclusion mentions increased computational cost, but no FLOPs, parameter counts, or runtime comparisons are reported; a quantitative complexity comparison would strengthen the paper.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: TransAdapter is an empirical architecture paper whose benchmark results are independent of its module definitions.

full rationale

This paper does not contain a derivation chain that reduces to its own inputs. The modules in Eqs. (1)-(8) are architectural proposals evaluated on standard UDA benchmarks, not quantities fitted to those benchmarks and then reported as predictions. No parameter is calibrated to a target accuracy and then claimed as a forecast. The pseudo-label self-training in Sec. 3.4 is self-referential learning, but it is a standard technique and not circular in the ledger sense. There are no load-bearing self-citations: the bibliography contains no prior works by Doruk, Oztop, or Ates, and no uniqueness theorem is imported from the authors' own prior work. The noticeable implementability concerns, such as the dimensional mismatch in Eq. (8) and the batch-level versus per-sample treatment of the GDD, are correctness or reproducibility issues rather than circularity, because they do not make any claimed result true by definition. The conclusion's acknowledged limitation about lacking task-specific adaptation modules is an honest scope statement, not a circular step. The central claims are therefore self-contained empirical contributions with no circular reasoning identified.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on several hand-chosen hyperparameters and two unvalidated assumptions about entropy and pseudo-label reliability. No new entities are introduced. The main empirical contribution depends on the correct implementation of the Cross Feature Transform, whose equations are internally inconsistent.

free parameters (3)
  • lambda_local = 0.1
    Weight for local adversarial loss, set manually for all tasks.
  • lambda_global = 0.01
    Weight for global adversarial loss, set manually for all tasks.
  • pseudo_label_confidence_threshold = not specified
    Threshold for filtering pseudo-labels in pixelwise transform; the paper says a threshold based on model accuracy but never gives the value.
assumptions (4)
  • domain assumption Source and target share the same label space
    Standard UDA assumption used throughout Section 3.
  • ad hoc to paper Entropy H(F_graph) is a reliable indicator of transferability
    Eq (2) multiplies attention by H, assuming high-entropy features are less transferable. The paper provides no external justification or measurement.
  • ad hoc to paper Pseudo-labels from a Swin-Base model trained only on source are accurate enough after confidence thresholding
    Section 3.4 uses these pseudo-labels to guide CutMix/MixUp; if they are wrong, the transformed source pairs teach incorrect alignments.
  • domain assumption ImageNet pretraining transfers to domain adaptation
    All models use ImageNet-pretrained Swin backbones; the paper relies on this without studying its effect.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TransAdapter: Vision Transformer for Feature-Centric Unsupervised Domain Adaptation." pith.science (2026). https://pith.science/paper/WFYQRMC4

@misc{pith2026241204073,
  author       = {Pith},
  title        = {Pith review of: TransAdapter: Vision Transformer for Feature-Centric Unsupervised Domain Adaptation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WFYQRMC4}},
  note         = {Machine review of arXiv:2412.04073}
}
read the original abstract

Unsupervised Domain Adaptation (UDA) aims to utilize labeled data from a source domain to solve tasks in an unlabeled target domain, often hindered by significant domain gaps. Traditional CNN-based methods struggle to fully capture complex domain relationships, motivating the shift to vision transformers like the Swin Transformer, which excel in modeling both local and global dependencies. In this work, we propose a novel UDA approach leveraging the Swin Transformer with three key modules. A Graph Domain Discriminator enhances domain alignment by capturing inter-pixel correlations through graph convolutions and entropy-based attention differentiation. An Adaptive Double Attention module combines Windows and Shifted Windows attention with dynamic reweighting to align long-range and local features effectively. Finally, a Cross-Feature Transform modifies Swin Transformer blocks to improve generalization across domains. Extensive benchmarks confirm the state-of-the-art performance of our versatile method, which requires no task-specific alignment modules, establishing its adaptability to diverse applications.

Figures

Figures reproduced from arXiv: 2412.04073 by the authors.

Figure 1
Figure 1. The architecture of the proposed TransAdapter; MADA is multi-head adaptive double attention module, [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The architecture of the Adaptive Double Attention (ADA) module is depicted. Here, [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The architecture of the Graph Domain Discriminator uses [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: The architecture of Cross Feature Transform (CFT) module. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: t-SNE visualization of Office-Home dataset, where red and blue points indicate the source and the target [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 40 canonical work pages

  1. [4]

    Large-scale machine learning with stochastic gradient descent

    [Bottou(2010)] Bottou, L. Large-scale machine learning with stochastic gradient descent. In Proceedings of COMP- STAT’2010, pp. 177–186. Springer,

    work page 2010

  2. [8]

    Domain adaptation for visual recognition: A comprehensive survey

    10 TransAdapter: Vision Transformer for Feature-Centric Unsupervised Domain Adaptation [Csurka(2017)] Csurka, G. Domain adaptation for visual recognition: A comprehensive survey. arXiv preprint arXiv:1702.05374,

    arXiv 2017

  3. [9]

    Imagenet: A large-scale hierarchical image database

    [Deng et al.(2009)Deng, Dong, Socher, Li, Li, and Fei-Fei] Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei- Fei, L. Imagenet: A large-scale hierarchical image database. 2009 IEEE conference on computer vision and pattern recognition, pp. 248–255,

    work page 2009

  4. [12]

    Domain-agnostic mutual prompting for unsupervised domain adaptation

    [Du et al.(2024)Du, Li, and Zhao] Du, X., Li, C., and Zhao, K. Domain-agnostic mutual prompting for unsupervised domain adaptation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2402–2410,

    work page 2024

  5. [13]

    and Lempitsky, V

    [Ganin & Lempitsky(2015)Ganin and Lempitsky] Ganin, Y . and Lempitsky, V . Unsupervised domain adaptation by backpropagation. In International conference on machine learning, pp. 1180–1189. PMLR,

    work page 2015

  6. [15]

    Deep residual learning for image recognition

    [He et al.(2016)He, Zhang, Ren, and Sun] He, K., Zhang, X., Ren, S., and Sun, J. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770–778,

    work page 2016

  7. [17]

    Energy-based self-training and normalization for unsupervised domain adaptation

    [Herath et al.(2023)Herath, Wang, and Huang] Herath, S., Wang, J., and Huang, J. Energy-based self-training and normalization for unsupervised domain adaptation. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp. 1620–1628,

    work page 2023

  8. [19]

    [Jiang et al.(2021)] Jiang, W. et al. Cross-domain learning with deep neural networks. Pattern Recognition, 115: 107888,

    work page 2021

Show all 52 references
  1. [20]

    [Jiang et al.(2022)] Jiang, Y . et al. Domain shift challenges in deep learning. Neural Computation, 34:1423–1442,

    2022

  2. [23]

    [Kuroki et al.(2019)] Kuroki, T. et al. An overview of unsupervised domain adaptation methods. Journal of AI Research, 17:45–67,

    2019

  3. [24]

    P., Li, X., Fu, K

    [Lai et al.(2023)Lai, Vesdapunt, Zhou, Wu, Huynh, Li, Fu, and Chuah] Lai, Z., Vesdapunt, N., Zhou, N., Wu, J., Huynh, C. P., Li, X., Fu, K. K., and Chuah, C.-N. Padclip: Pseudo-labeling with adaptive debiasing in clip for unsupervised domain adaptation. In Proceedings of the I...

    2023

  4. [25]

    Drop to adapt: Learning discriminative features for unsupervised domain adaptation

    [Lee et al.(2019)Lee, Kim, Kim, and Jeong] Lee, S., Kim, D., Kim, N., and Jeong, S.-G. Drop to adapt: Learning discriminative features for unsupervised domain adaptation. In Proceedings of the IEEE/CVF international conference on computer vision, pp. 91–100,

    2019

  5. [26]

    [Li et al.(2021)] Li, Y . et al. Towards robust domain adaptation.IEEE Transactions on Neural Networks and Learning Systems, 32:569–583,

    2021

  6. [27]

    Do we really need to access the source data? source hypothesis transfer for unsupervised domain adaptation

    [Liang et al.(2020)Liang, Hu, and Feng] Liang, J., Hu, D., and Feng, J. Do we really need to access the source data? source hypothesis transfer for unsupervised domain adaptation. In International conference on machine learning, pp. 6028–6039. PMLR,

    2020

  7. [28]

    Transferable adversarial training: A general approach to adapting deep classifiers

    [Liu et al.(2019)Liu, Long, Wang, and Jordan] Liu, H., Long, M., Wang, J., and Jordan, M. Transferable adversarial training: A general approach to adapting deep classifiers. In International conference on machine learning, pp. 4013–4022. PMLR,

    2019

  8. [29]

    Swin transformer: Hierarchical vision transformer using shifted windows

    [Liu et al.(2021a)Liu, Lin, Cao, Hu, Wei, Zhang, Lin, and Guo] Liu, Z., Lin, Y ., Cao, Y ., Hu, H., Wei, Y ., Zhang, Z., Lin, S., and Guo, B. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF International Conference on Com...

    2021 arXiv

  9. [30]

    H., Wang, Y ., and Li, W

    [Ma et al.(2021)Ma, Zhang, Li, Liu, Wang, and Li] Ma, W., Zhang, J., Li, S., Liu, C. H., Wang, Y ., and Li, W. Exploit- ing both domain-specific and invariant knowledge via a win-win transformer for unsupervised domain adaptation. arXiv preprint arXiv:2111.12941,

    2021 arXiv

  10. [31]

    [Morerio et al.(2020)] Morerio, P. et al. Noisy labels in domain adaptation: A robust approach. IEEE Transactions on Pattern Analysis and Machine Intelligence, 42:2051–2064,

    2020

  11. [32]

    [Oza et al.(2021)] Oza, P. et al. Efficient domain adaptation with deep learning. IEEE Transactions on Cybernetics, 51: 812–825,

    2021

  12. [33]

    Visda: The visual domain adaptation challenge

    [Peng et al.(2017)Peng, Usman, Kaushik, Hoffman, Wang, and Saenko] Peng, X., Usman, B., Kaushik, N., Hoffman, J., Wang, D., and Saenko, K. Visda: The visual domain adaptation challenge. In arXiv preprint arXiv:1710.06924,

    2017 arXiv

  13. [34]

    Y ., Saenko, K., and Wang, B

    [Peng et al.(2019)Peng, Bai, Xia, Huang, Saenko, and Wang] Peng, X., Bai, Z., Xia, X., Huang, Z. Y ., Saenko, K., and Wang, B. Moment matching for multi-source domain adaptation. In Proceedings of the IEEE International Conference on Computer Vision (ICCV), pp. 1406–1415,

    2019

  14. [35]

    [Qian et al.(2021)] Qian, Y . et al. Recent advancements in deep neural networks for computer vision.IEEE Transactions on Pattern Analysis and Machine Intelligence, 43(8):2456–2473,

    2021

  15. [36]

    Adapting visual category models to new domains

    [Saenko et al.(2010)Saenko, Kulis, Fritz, and Darrell] Saenko, K., Kulis, B., Fritz, M., and Darrell, T. Adapting visual category models to new domains. In European Conference on Computer Vision, pp. 213–226. Springer,

    2010

  16. [37]

    Max- imum classifier discrepancy for unsupervised domain adaptation

    [Saito et al.(2018)Saito, Watanabe, Ushiku, and Harada] Saito, K., Watanabe, K., Ushiku, Y ., and Harada, T. Max- imum classifier discrepancy for unsupervised domain adaptation. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 3723–3732,

    2018

  17. [38]

    Domain-specificity inducing trans- formers for source-free domain adaptation

    [Sanyal et al.(2023)Sanyal, Gupta, and Roy] Sanyal, R., Gupta, A., and Roy, P. Domain-specificity inducing trans- formers for source-free domain adaptation. arXiv preprint arXiv:2308.14023,

    2023 arXiv

  18. [39]

    Safe self-refinement for transformer-based domain adaptation

    [Sun et al.(2022)Sun, Lu, Zhang, and Ling] Sun, T., Lu, C., Zhang, T., and Ling, H. Safe self-refinement for transformer-based domain adaptation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 7191–7200,

    2022

  19. [40]

    [Tan et al.(2019)] Tan, M. et al. A survey on deep learning for computer vision applications. Neurocomputing, 354: 30–44,

    2019

  20. [41]

    Training data-efficient image transformers & distillation through attention

    [Touvron et al.(2021)Touvron, Cord, Douze, Massa, Sablayrolles, and Jégou] Touvron, H., Cord, M., Douze, M., Massa, F., Sablayrolles, A., and Jégou, H. Training data-efficient image transformers & distillation through attention. In International conference on machine learning,...

    2021

  21. [42]

    Deep domain confusion: Maximizing for domain invariance

    [Tzeng et al.(2014a)Tzeng, Hoffman, Saenko, and Darrell] Tzeng, E., Hoffman, J., Saenko, K., and Darrell, T. Deep domain confusion: Maximizing for domain invariance. In arXiv preprint arXiv:1412.3474, 2014a. [Tzeng et al.(2014b)Tzeng, Hoffman, Zhang, Saenko, and Darrell] Tzeng...

    2014 arXiv

  22. [43]

    Deep hashing network for unsupervised domain adaptation

    [Venkateswara et al.(2017)Venkateswara, Eusebio, Chakraborty, and Panchanathan] Venkateswara, H., Eusebio, J., Chakraborty, S., and Panchanathan, S. Deep hashing network for unsupervised domain adaptation. In Pro- ceedings of the IEEE Conference on Computer Vision and Pattern ...

    2017

  23. [44]

    [VS et al.(2021)] VS, R. et al. A comprehensive survey on unsupervised domain adaptation. IEEE Transactions on Neural Networks and Learning Systems, 32:800–818,

    2021

  24. [45]

    Domain adaptation via bidirectional cross-attention transformer

    12 TransAdapter: Vision Transformer for Feature-Centric Unsupervised Domain Adaptation [Wang et al.(2022a)Wang, Guo, and Zhang] Wang, X., Guo, P., and Zhang, Y . Domain adaptation via bidirectional cross-attention transformer. arXiv preprint arXiv:2201.05887, 2022a. [Wang et a...

    2022 arXiv

  25. [46]

    Cdtrans: Cross-domain transformer for unsupervised domain adaptation

    [Xu et al.(2021)Xu, Chen, Wang, Wang, Li, and Jin] Xu, T., Chen, W., Wang, P., Wang, F., Li, H., and Jin, R. Cdtrans: Cross-domain transformer for unsupervised domain adaptation. arXiv preprint arXiv:2109.06165,

    2021 arXiv

  26. [47]

    Tvt: Transferable vision trans- former for unsupervised domain adaptation

    [Yang et al.(2023)Yang, Liu, Xu, and Huang] Yang, J., Liu, J., Xu, N., and Huang, J. Tvt: Transferable vision trans- former for unsupervised domain adaptation. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pp. 520–530,

    2023

  27. [48]

    J., Chun, S., Choe, J., and Yoo, Y

    [Yun et al.(2019)Yun, Han, Oh, Chun, Choe, and Yoo] Yun, S., Han, D., Oh, S. J., Chun, S., Choe, J., and Yoo, Y . Cutmix: Regularization strategy to train strong classifiers with localizable features. In Proceedings of the IEEE/CVF international conference on computer vision, ...

    2019

  28. [49]

    mixup: Beyond empirical risk minimization

    [Zhang(2017)] Zhang, H. mixup: Beyond empirical risk minimization. arXiv preprint arXiv:1710.09412,

    2017 arXiv

  29. [50]

    [Zhang et al.(2019)] Zhang, W. et al. Category-level domain adaptation for neural networks. IEEE Transactions on Neural Networks and Learning Systems, 30:277–287,

    2019

  30. [51]

    [Zhang et al.(2020)] Zhang, W. et al. Learning transferable features with deep adaptation networks. IEEE Transactions on Neural Networks and Learning Systems, 31:812–825,

    2020

  31. [52]

    [Zhao et al.(2020)] Zhao, H. et al. Domain adaptation challenges in deep neural networks. IEEE Transactions on Neural Networks and Learning Systems, 31:2031–2043,

    2020

  32. [53]

    Patch-mix transformer for unsupervised domain adaptation: A game perspective

    [Zhu et al.(2023)Zhu, Bai, and Wang] Zhu, J., Bai, H., and Wang, L. Patch-mix transformer for unsupervised domain adaptation: A game perspective. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 3561–3571,

    2023

  33. [2009]

    [Devlin et al.(2018)] Devlin, J. et al. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805,

    2018 arXiv

  34. [2010]

    [Bousmalis et al.(2017)] Bousmalis, K. et al. Unsupervised domain adaptation by backpropagation.Journal of Machine Learning Research, 18:1–35,

    2017

  35. [2015]

    [Han et al.(2020)] Han, K. et al. Generative adversarial transformers. arXiv preprint arXiv:2008.07772,

    2020 arXiv

  36. [2016]

    [He et al.(2021)] He, K. et al. Masked autoencoders are scalable vision learners. arXiv preprint arXiv:2111.06377,

    2021 arXiv

  37. [2017]

    [Chen et al.(2021)] Chen, L. et al. Understanding domain adaptation in deep learning. Artificial Intelligence Review, 54:1265–1282,

    2021

  38. [2018]

    [Dosovitskiy et al.(2020)] Dosovitskiy, A. et al. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929,

    2020 arXiv

  39. [2019]

    [Khan et al.(2021)] Khan, M. et al. Transformers in vision: A survey. ACM Computing Surveys, 54:1–41,

    2021

  40. [2020]

    A comprehensive survey on vision transformers in domain adaptation and domain generalization

    [Alijani et al.(2024)Alijani, Zhou, and Wang] Alijani, S., Zhou, Y ., and Wang, M. A comprehensive survey on vision transformers in domain adaptation and domain generalization. arXiv preprint arXiv:2404.04452,

    2024 arXiv

  41. [2021]

    Adversarial-learned loss for domain adaptation

    [Chen et al.(2020)Chen, Zhao, Liu, and Cai] Chen, M., Zhao, S., Liu, H., and Cai, D. Adversarial-learned loss for domain adaptation. In Proceedings of the AAAI conference on artificial intelligence, volume 34, pp. 3521–3528,

    2020

  42. [2022]

    [Kang et al.(2019)] Kang, G. et al. Deep adaptation networks for domain adaptation. Journal of Machine Learning Research, 20:1–30,

    2019

  43. [2023]

    [Jiang et al.(2020)] Jiang, W. et al. Improving domain adaptation with pseudo-labels.Pattern Recognition, 104:107248,

    2020

  44. [2024]

    Prompt-based distribution alignment for unsupervised domain adaptation

    [Bai et al.(2024)Bai, Zhang, Zhou, Huang, Luan, Wang, and Chen] Bai, S., Zhang, M., Zhou, W., Huang, S., Luan, Z., Wang, D., and Chen, B. Prompt-based distribution alignment for unsupervised domain adaptation. In Proceedings of the AAAI Conference on Artificial Intelligence, v...

    2024

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.