Attract or Distract: Exploit the Margin of Open Set

Guoliang Kang; Hehe Fan; Qianyu Feng; Yi Yang

REVIEW 3 major objections 6 minor 43 references

Attract or Distract: Exploit the Margin of Open Set

T0 review · 3 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Exploiting semantic structure—aligning target class centroids to source centroids and pushing unknown samples beyond an adaptive margin—improves open set domain adaptation, reaching 84.3% on Digits and 89.7% OS on Office-31.

desk verdict A plausible open-set domain adaptation method whose only pseudo-label safeguard is vacuous; the empirical gains are real but the mechanism is unproven. read the letter →

arxiv 1908.01925 v2 pith:MSTLKT2M submitted 2019-08-06 cs.CV

Qianyu Feng , Guoliang Kang , Hehe Fan , Yi Yang This is my paper · ORCID

classification cs.CV

keywords opensetdomainadaptationsemanticcategoricalalignmentcontrastivemappingadaptivemargincontrastive-centerlossunknownclassdetectioncentroidadversarial

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

Open set domain adaptation asks a model trained on labeled source classes to classify target data that includes both shared classes and an unknown class never seen in the source. The paper tries to establish that the semantic structure of the open set data is the missing ingredient: known classes should be compact and aligned across domains, while unknown samples should be pushed away from all known-class centroids. It proposes two losses, Semantic Categorical Alignment (SCA) and Semantic Contrastive Mapping (SCM), built on top of adversarial domain adaptation, and reports that they improve average accuracy on Digit benchmarks to 84.3% and OS accuracy on Office-31 with AlexNet to 89.7%. A sympathetic reader would take the central claim to be that explicitly enlarging two margins—between known classes and between unknown and known classes—makes the open set decision boundary more reliable.

What carries the argument

The load-bearing objects are class centroids $c^k_s$ and $c^k_t$ in the feature space, updated per iteration with a cosine-similarity reweighting that down-weights faraway mini-batch centroids. Three losses run together: adversarial domain adaptation aligns source and target distributions while treating unknown as class $N+1$; the contrastive-center loss $L_{cct}$ compacts source samples; and the categorical center alignment loss $L_{cca}$ ties each target centroid to its source counterpart. The Semantic Contrastive Mapping uses distance functions from energy-based contrastive learning: known target samples are pulled to source centroids, unknown samples are penalized when they fall inside the adaptive margin $M^k$ of a known class. The adaptive margin is what lets the model grow the unknown/known separation as alignment proceeds.

What would settle it

Train the model while randomly corrupting the pseudo-labels of target known samples at a fixed rate, then compare final OS/OS* accuracy and the distance between estimated and true target centroids. If SCA/SCM still improve over the adversarial baseline under strong corruption, the losses are robust; if performance collapses, the reported gains depend on pseudo-label accuracy.

Watch

Extended reading notes

Core claim

The central claim is that the open set problem is best handled by making the margin structure explicit. SCA computes class centroids in source and target feature spaces, compacts source samples around their own centroid with a contrastive-center loss, and aligns each target centroid to its corresponding source centroid through a cosine-similarity reweighted update. SCM then selects target samples whose predicted probability passes a threshold, attracts known samples to their source centroid, and repels unknown samples from every known centroid using a class-dependent adaptive margin $M^k$ given by the average distance from the target centroid of class $k$ to the other source centroids. The paper argues that this combination produces more discriminative features, better cross-domain centroid alignment, and a clearer unknown/known boundary, and supports the claim with ablations, feature visualizations, and accuracy gains over the OSDA+BP baseline.

Load-bearing premise

The method assumes the model's own guesses about target class labels are trustworthy enough for estimating centroids and selecting reliable samples; when those guesses are systematically wrong, the alignment and distraction losses can reinforce the error instead of correcting it.

Editorial extensions

If this is right

On the Digit benchmarks, the full method reaches 84.3% average OS accuracy, 1.9% higher than the prior best, and raises unknown-class accuracy to 87.9%.
On Office-31 with AlexNet, the method reaches 89.7% OS and 90.7% OS*, improving over the OSDA+BP baseline by 2% and 10.5% respectively.
Ablation studies show that removing either SCA or SCM lowers accuracy, so the two losses work jointly rather than one carrying the whole improvement.
Experiments with a static margin degrade accuracy, indicating the unknown/known margin must expand as cross-domain alignment shrinks inter-class distances.

Reading between the lines

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

The paper leaves implicit that the pseudo-label threshold could be made adaptive to per-class confidence; a testable extension is to replace the fixed $1/(N+1)$ cutoff with a calibrated uncertainty estimate.
Because the mechanism only needs class centroids and an adaptive margin, the same attract/distract idea could be lifted to open-set recognition without a labeled source, treating learned prototypes as centroids.
A direct stress test would vary the amount of pseudo-label noise on target known classes and check whether SCA/SCM still beats the adversarial baseline; the paper asserts robustness but does not bound the noise level at which the gains disappear.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Two things you should know. First, this is a useful combination of known ingredients: contrastive-center loss for source compactness, centroid alignment across domains, and a contrastive loss with an adaptive margin to push unknown target samples away. On Digits and Office-31 it beats OSDA+BP by 2–3 points on average, and the ablations show each module contributes. Second, the paper's only stated defense against pseudo-label noise is a threshold that literally does nothing. The threshold is set to 1/(N+1); for a softmax over N+1 classes the max probability is always at least that value, so reliable-sample selection excludes only exact ties. Every confident but wrong sample sails through. The centroids in Eqs. (5)–(7) and the contrastive mapping in Eq. (9) are built from those pseudo-labels, so the central mechanism is fed by unvetted noise. The paper asserts in Section 1 that 'SCA and SCM in itself are robust to such noisy labels,' but gives no bound, no pseudo-label accuracy measurement, and no experiment with oracle labels.

What's genuinely good: the adaptive margin in Eq. (12) is a sensible idea—using the average distance to other source centroids as a per-class radius. The cosine reweighting is also reasonable. The evaluation is fairly thorough for the time: three digit tasks, Office-31 with two backbones, ablations, and a sensitivity study for omega. They report averages over three runs, though no error bars.

Where it's soft: beyond the threshold, the hyperparameters (lambda_s, lambda_c, lambda_t, omega) are selected on the test benchmarks; the omega grid search is on A->D and then applied everywhere. No code is released. And the Office-31 ablations show that w/o SCA or w/o SCM already beats OSDA+BP by large margins (e.g., A->D 87.8 vs 76.6 and 89.8 vs 76.6), so the incremental gain of the full method over its own components is small and could be within run-to-run variation. The paper doesn't report significance.

Net: is the central claim true? It might be, but the current evidence doesn't establish the mechanism. A skeptical reader could argue the gains come from the combination of adversarial training plus any of the extra losses, not from robust pseudo-label exploitation. The paper deserves a serious referee—the problem is real, the method is clearly described, and the results are positive—but it needs major revision. The easiest fix is to show pseudo-label accuracy during training, compare the vacuous threshold against stricter ones, and add an oracle-label upper bound. Without that, the robustness claim is just an assertion.

Who this is for: people doing open set domain adaptation or pseudo-label-based domain adaptation. I'd bring it to a reading group as a cautionary case in threshold design. My recommendation: engage with it, but push the authors to make the pseudo-label quality story concrete.

Referee Report

3 major / 6 minor

Summary. The paper proposes an open-set domain adaptation (OSDA) method in which only source labels are available and the target domain contains an unknown class not present in the source. The method combines adversarial domain adaptation with two semantic-structure losses: Semantic Categorical Alignment (SCA), which applies a contrastive-center loss on source features and aligns source/target class centroids, and Semantic Contrastive Mapping (SCM), which pulls target samples predicted as known toward the corresponding source centroids and pushes predicted-unknown samples away from all known centroids using an adaptive margin. Target pseudo-labels are used to estimate centroids and to select 'reliable' samples. Experiments on Digit datasets (SVHN, MNIST, USPS) and Office-31 (AlexNet/VGG) report state-of-the-art or competitive OS/OS* accuracy, with ablations showing that removing either module generally degrades performance.

Significance. If the reported results are reproducible, the paper makes a useful empirical contribution: it demonstrates that categorical centroid alignment and contrastive mapping, combined with a standard adversarial OSDA baseline, improve separation of known and unknown classes, and the ablations support that both modules contribute. The paper also provides visualizations and sensitivity analyses for the margin and re-weighting parameters. The main weakness is that the target-side supervision rests on pseudo-labels whose reliability is asserted but not demonstrated; the proposed 'reliable sample selection' is, as written, vacuous. Because the method's central mechanism depends on these pseudo-labels, the empirical claims need to be re-supported by either a corrected selection criterion or an explicit noise-robustness analysis.

major comments (3)

[Section 3.4, Eqs. (9)-(12)] The reliable sample selection is vacuous. With a softmax classifier over N+1 classes, every sample's maximum predicted probability is at least 1/(N+1), so the threshold 1/(N+1) in Section 3.4 excludes only exact uniform ties and admits essentially all target samples. The pseudo-label errors therefore flow into the target centroids in Eqs. (5)-(7) and into the contrastive loss in Eq. (9), including exactly the confident errors that are most damaging. The claim in Section 1 that SCA/SCM are 'robust to such noisy labels' is unsupported: no noise bound, no pseudo-label accuracy measurement, and no experiment with corrupted pseudo-labels is provided. This is load-bearing because the target-side terms are the paper's main novelty. Please either use a threshold that actually removes low-confidence samples, or add an explicit robustness analysis and experiment demonstrating that the method tolerates pseudo-label noise.
[Section 4.1, Fig. 4(a)] The hyper-parameters are selected on the evaluation benchmarks without a held-out validation set. The text fixes lambda_s=0.02, lambda_c=0.005, lambda_t=1e-4 and reports a sweep over omega in Fig. 4(a) (the scenario is not identified) that selects omega=0.5, the value used for all reported results. Since the improvements over the state of the art are small (e.g., 1.9% average on Digits), the absence of error bars (Tables 1-2 report only means over three runs) and the absence of a validation protocol make it difficult to assess whether the gain is genuine or a consequence of test-set tuning. Please provide standard deviations and either a validation split or a full sensitivity analysis showing that the ranking of methods is stable across the hyper-parameter range.
[Section 3.4, Eq. (12)] The adaptive margin M^k in Eq. (12) inherits pseudo-label bias because it is computed from target centroids c_j^t obtained from the model's own predictions. The cosine re-weighting in Eqs. (6)-(7) and (10)-(11) does not address the main failure mode: a sample from class j that is confidently but incorrectly predicted as class k is close to the source centroid of k, receives a high rho, and therefore updates c_k^t toward class j. This can reinforce the error in both the centroid alignment and the SCM margin. Please add a discussion of this failure mode and, ideally, a controlled experiment (e.g., injecting known pseudo-label corruption and measuring the effect on OS/OS*) to support the robustness claim.

minor comments (6)

[Section 3.3, Eq. (4)] The text says 'delta = 1 is set to be 10^{-6}'. This is contradictory; it should read 'delta is set to 10^{-6}'.
[Section 3.2, Eq. (1)] Equation (1) has a typo: the closing parenthesis in '- log(D∘G(xs))ys )' is misplaced; the subscript should be on the probability, not outside the parentheses.
[Section 3.2, Eq. (3)] The min/max structure of Eq. (3) is ambiguous: L_adv appears with a minus sign under min_G and a plus sign under min_D, but the overall equality with a sum of two minima does not clearly specify how the gradient-reversal layer interacts with the two terms. Please write the objective as a single min-max expression for clarity.
[Algorithm 1] The loop structure is unclear: the outer loop increments e while the inner loop runs to max_iter, but e is never used in the updates. Please clarify whether the outer loop is intended to repeat over epochs and where the centroids are re-initialized.
[References] Reference [4] is incomplete: 'Q. Ce and S. Fei' should be expanded to the actual author names and full publication details; the current form makes it hard to verify the contrastive-center loss.
[Table 1] In Table 1, for USPS→MNIST the variant 'w/o SCA' yields a higher OS score than the full method (93.6 vs 93.1). The text says removing SCA decreases performance, which is not uniformly true across all metrics and tasks. Please qualify the ablation claim accordingly.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported OS/OS* gains are evaluated on held-out target labels, and the pseudo-label-based training losses are not a fitted re-statement of the reported metric.

full rationale

Walking the derivation chain, the central claim is supported by Table 1 and Table 2, which report OS/OS* accuracies computed against ground-truth target labels. The training objective (Eq. 13) uses network predictions as pseudo-labels for target centroids (Eqs. 5-7), reliable-sample selection (Section 3.4), and the contrastive mapping (Eq. 9), but these are optimization inputs, not the evaluation quantity. No equation defines the final accuracy in terms of the loss components, and no fitted parameter is later renamed as a prediction. The self-citations (e.g., [7], [8], [18], [19], [42], [43]) appear only in related-work enumerations and do not carry a load-bearing uniqueness claim, ansatz, or definitional premise. The threshold 1/(N+1) in Section 3.4 is admittedly weak as a reliability filter, but this is a pseudo-label-noise robustness concern rather than a circular reduction, because the reported numbers are not guaranteed by construction. The method is also benchmarked against external baselines (OSVM, MMD+OSVM, BP+OSVM, OSDA+BP), so the empirical claim is self-contained and not circular.

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

The paper rests on a few standard assumptions: unknown classes can be collapsed into one class, pseudo-labels from the model are reliable enough for centroid estimation, and the contrastive/center losses improve feature structure. The main free parameters are loss weights and the reweighting exponent, which are set or tuned on the benchmark without a separate validation protocol.

free parameters (4)

reweighting exponent omega = 0.5
Swept over {0, 0.1, 0.5, 1, 1.5, 2} in Fig 4(a) and the best value is chosen for all experiments. This is a free parameter tuned on the target benchmark.
loss weights lambda_s, lambda_c, lambda_t = 0.02, 0.005, 1e-4
Set to fixed values for all experiments, but no validation protocol is described; they are likely chosen on the benchmarks.
reliable sample threshold = 1/(N+1)
Threshold for selecting reliable target samples based on classification probability; chosen by the authors without a sensitivity analysis.
delta (denominator stabilizer) = 1e-6
Constant in Eq. (4) to prevent zero denominator; a standard stabilizer and not a meaningful free parameter.

assumptions (4)

domain assumption The unknown class can be treated as a single class in the target domain
The paper assumes all unshared target classes can be lumped into one 'unknown' class, following Saito et al. [22]. This is a standard OSDA assumption but may not hold when the unknown classes are multiple and diverse.
domain assumption Target pseudo-labels, obtained from the model itself, are reliable enough to estimate centroids and drive the SCA/SCM losses
The method uses model predictions as pseudo-labels for target samples. The correctness depends on this self-training loop not diverging; the paper claims robustness empirically but provides no formal analysis.
standard math The contrastive-center loss improves feature discriminability and thus helps alignment
Borrowed from [4] and used in CAN [18]; it is a known technique that typically improves intra-class compactness and inter-class separability.
domain assumption Adversarial domain adaptation (OSDA+BP) provides a reasonable base feature space
The ADA module is taken from [22]; the overall method inherits its assumptions about the discriminator and gradient reversal.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Attract or Distract: Exploit the Margin of Open Set." pith.science (2026). https://pith.science/paper/MSTLKT2M

@misc{pith2026190801925,
  author       = {Pith},
  title        = {Pith review of: Attract or Distract: Exploit the Margin of Open Set},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MSTLKT2M}},
  note         = {Machine review of arXiv:1908.01925}
}

read the original abstract

Open set domain adaptation aims to diminish the domain shift across domains, with partially shared classes. There exist unknown target samples out of the knowledge of source domain. Compared to the close set setting, how to separate the unknown (unshared) class from the known (shared) ones plays a key role. Whereas, previous methods did not emphasize the semantic structure of the open set data, which may introduce bias into the domain alignment and confuse the classifier around the decision boundary. In this paper, we exploit the semantic structure of open set data from two aspects: 1) Semantic Categorical Alignment, which aims to achieve good separability of target known classes by categorically aligning the centroid of target with the source. 2)Semantic Contrastive Mapping, which aims to push the unknown class away from the decision boundary. Empirically, we demonstrate that our method performs favourably against the state-of-the-art methods on representative benchmarks, e.g. Digit datasets and Office-31 datasets.

Figures

Figures reproduced from arXiv: 1908.01925 by the authors.

**Figure 2.** Framework of the proposed method. There are three modules: Adversarial Domain Adaptation (ADA), Semantic Categorical [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** A comparison between the existing method and the proposed method. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: (a): A comparison of the behavior of our method with different re-weight ω in the contrastive loss. (b): A comparison between static margin and our adaptive margin. (c): Distances between centroid of class “backpack” labeled “0” in the target with centroids of class 0∼…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 40 canonical work pages

[1]

Baktashmotlagh, M

M. Baktashmotlagh, M. Faraki, T. Drummond, and M. Salz- mann. Learning factorized representations for open-set do- main adaptation. In ICLR, 2019. 1, 3

work page 2019
[2]

Bousmalis, N

K. Bousmalis, N. Silberman, D. Dohan, D. Erhan, and D. Kr- ishnan. Unsupervised pixel-level domain adaptation with generative adversarial networks. In CVPR, 2017. 2

work page 2017
[3]

Busto and J

P. Busto and J. Gall. Open set domain adaptation. In ICCV,

work page
[4]

Ce and S

Q. Ce and S. Fei. Contrastive-center loss for deep neural networks. In 2017 IEEE International Conference on Image Processing (ICIP), pages 2851–2855, 2017. 2, 4

work page 2017
[5]

B. Chen, W. Deng, and H. Shen. Virtual class enhanced dis- criminative embedding learning. In Advances in Neural In- formation Processing Systems, pages 1942–1952, 2018. 2

work page 1942
[6]

J. Deng, W. Dong, R. Socher, L. Li, K. Li, and L. Fei-Fei. Im- agenet: A large-scale hierarchical image database. In CVPR,

work page
[7]

H. Fan, X. Chang, D. Cheng, Y . Yang, D. Xu, and A. G. Hauptmann. Complex event detection by identifying reliable shots from untrimmed videos. In ICCV, 2017. 2

work page 2017
[8]

H. Fan, L. Zheng, C. Yan, and Y . Yang. Unsupervised person re-identiﬁcation: Clustering and ﬁne-tuning. ACM Transac- tions on Multimedia Computing, Communications, and Ap- plications (TOMM), 14(4):83, 2018. 2

work page 2018

Show all 43 references

[9]

Ganin and V

Y . Ganin and V . Lempitsky. Unsupervised domain adaptation by backpropagation. In ICML, 2015. 2, 4, 5, 6

2015
[10]

Ghifary, W

M. Ghifary, W. B. Kleijn, M. Zhang, D. Balduzzi, and W. Li. Deep reconstruction-classiﬁcation networks for un- supervised domain adaptation. In ECCV, 2016. 2

2016
[11]

Gretton, K

A. Gretton, K. Borgwardt, M. Rasch, B. Sch ¨olkopf, and A. J. Smola. A kernel method for the two-sample-problem. In NIPS, pages 513–520, 2007. 5

2007
[12]

Guan and Y

Q. Guan and Y . Huang. Multi-label chest x-ray image classi- ﬁcation via category-wise residual attention learning.Pattern Recognition Letters, 2018. 2

2018
[13]

Hadsell, S

R. Hadsell, S. Chopra, and Y . LeCun. Dimensionality re- duction by learning an invariant mapping. In CVPR, 2006. 5

2006
[14]

Haeusser, T

P. Haeusser, T. Frerix, A. Mordvintsev, and D. Cremers. As- sociative domain adaptation. In ICCV, 2017. 2

2017
[15]

K. He, G. Gkioxari, P. Doll ´ar, and R. Girshick. Mask r-cnn. In ICCV, 2017. 1

2017
[16]

K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In CVPR, 2016. 1

2016
[17]

L. Jain, W. Scheirer, and T. Boult. Multi-class open set recognition using probability of inclusion. In ECCV, 2014. 5

2014
[18]

G. Kang, L. Jiang, Y . Yang, and A. G. Hauptmann. Con- trastive adaptation network for unsupervised domain adapta- tion. In CVPR, 2019. 2

2019
[19]

G. Kang, L. Zheng, Y . Yan, and Y . Yang. Deep adversarial attention alignment for unsupervised domain adaptation: the beneﬁt of target expectation maximization. In ECCV, 2018. 2

2018
[20]

D. P. Kingma and J. Ba. Adam: A method for stochastic optimization. In ICLR, 2015. 6

2015
[21]

Krizhevsky, I

A. Krizhevsky, I. Sutskever, and G. E. Hinton. Imagenet classiﬁcation with deep convolutional neural networks. In NIPS, pages 1097–1105, 2012. 1, 6

2012
[22]

Kuniaki, Y

S. Kuniaki, Y . Shohei, U. Yoshitaka, and H. Tatsuya. Open set domain adaptation by backpropagation. In ECCV, 2018. 1, 2, 4, 5, 6, 7

2018
[23]

LeCun, L

Y . LeCun, L. Bottou, Y . Bengio, P. Haffner, et al. Gradient- based learning applied to document recognition. Proceed- ings of the IEEE, 86(11):2278–2324, 1998. 6

1998
[24]

J. Long, E. Shelhamer, and T. Darrell. Fully convolutional networks for semantic segmentation. In CVPR, 2015. 1

2015
[25]

M. Long, Y . Cao, J. Wang, and M. I. Jordan. Learning trans- ferable features with deep adaptation networks. In ICML,
[26]

M. Long, H. Zhu, J. Wang, and M. I. Jordan. Unsuper- vised domain adaptation with residual transfer networks. In Advances in Neural Information Processing Systems , pages 136–144, 2016. 2

2016
[27]

Y . Luo, P. Liu, T. Guan, J. Yu, and Y . Yang. Signiﬁcance- aware information bottleneck for domain adaptive semantic segmentation. In ICCV, 2019. 2

2019
[28]

Y . Luo, L. Zheng, T. Guan, J. Yu, and Y . Yang. Taking a closer look at domain shift: Category-level adversaries for semantics consistent domain adaptation. In CVPR, 2019. 2

2019
[29]

Netzer, T

Y . Netzer, T. Wang, A. Coates, A. Bissacco, B. Wu, and A. Ng. Reading digits in natural images with unsupervised feature learning. In NIPS, 2011. 6

2011
[30]

Redmon, S

J. Redmon, S. Divvala, R. Girshick, and A. Farhadi. You only look once: Uniﬁed, real-time object detection. In CVPR, 2016. 1

2016
[31]

Saenko, B

K. Saenko, B. Kulis, M. Fritz, and T. Darrell. Adapting vi- sual category models to new domains. In ECCV, 2010. 5, 6

2010
[32]

Saito, Y

K. Saito, Y . Ushiku, and T. Harada. Asymmetric tri-training for unsupervised domain adaptation. In ICML, 2017. 2

2017
[33]

Saito, Y

K. Saito, Y . Ushiku, T. Harada, and K. Saenko. Adversarial dropout regularization. In ICLR, 2018. 2

2018
[34]

Saito, K

K. Saito, K. Watanabe, Y . Ushiku, and T. Harada. Maximum classiﬁer discrepancy for unsupervised domain adaptation. In CVPR, 2018. 2

2018
[35]

Sener, H

O. Sener, H. O. Song, A. Saxena, and S. Savarese. Learning transferrable representations for unsupervised domain adap- tation. In Advances in Neural Information Processing Sys- tems, pages 2110–2118, 2016. 2

2016
[36]

R. Shu, H. H. Bui, H. Narui, and S. Ermon. A dirt-t approach to unsupervised domain adaptation. ICLR, 2018. 2

2018
[37]

Simonyan and A

K. Simonyan and A. Zisserman. Very deep convolutional networks for large-scale image recognition. In ICLR, 2015. 1, 6

2015
[38]

Szegedy, W

C. Szegedy, W. Liu, Y . Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V . Vanhoucke, and A. Rabinovish. Going deeper with convolutions. In CVPR, 2015. 1

2015
[39]

Tzeng, J

E. Tzeng, J. Hoffman, K. Saenko, and T. Darrell. Adversarial discriminative domain adaptation. In CVPR, 2017. 2

2017
[40]

Tzeng, J

E. Tzeng, J. Hoffman, N. Zhang, K. Saenko, and T. Darrell. Deep domain confusion: Maximizing for domain invariance. arXiv preprint arXiv:1412.3474, 2014. 2

2014 arXiv
[41]

Zheng, L

Z. Zheng, L. Zheng, M. Garrett, Y . Yang, and Y .-D. Shen. Dual-path convolutional image-text embedding with instance loss. arXiv preprint arXiv:1711.05535, 2017. 2

2017 arXiv
[42]

Zhong, L

Z. Zhong, L. Zheng, Z. Luo, S. Li, and Y . Yang. Invariance matters: Exemplar memory for domain adaptive person re- identiﬁcation. In CVPR, 2019. 2

2019
[43]

Zhong, L

Z. Zhong, L. Zheng, Z. Zheng, S. Li, and Y . Yang. Cam- style: A novel data augmentation method for person re- identiﬁcation. IEEE Transactions on Image Processing , 28:1176–1190, 2019. 2

2019

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Baktashmotlagh, M

M. Baktashmotlagh, M. Faraki, T. Drummond, and M. Salz- mann. Learning factorized representations for open-set do- main adaptation. In ICLR, 2019. 1, 3

work page 2019

[2] [2]

Bousmalis, N

K. Bousmalis, N. Silberman, D. Dohan, D. Erhan, and D. Kr- ishnan. Unsupervised pixel-level domain adaptation with generative adversarial networks. In CVPR, 2017. 2

work page 2017

[3] [3]

Busto and J

P. Busto and J. Gall. Open set domain adaptation. In ICCV,

work page

[4] [4]

Ce and S

Q. Ce and S. Fei. Contrastive-center loss for deep neural networks. In 2017 IEEE International Conference on Image Processing (ICIP), pages 2851–2855, 2017. 2, 4

work page 2017

[5] [5]

B. Chen, W. Deng, and H. Shen. Virtual class enhanced dis- criminative embedding learning. In Advances in Neural In- formation Processing Systems, pages 1942–1952, 2018. 2

work page 1942

[6] [6]

J. Deng, W. Dong, R. Socher, L. Li, K. Li, and L. Fei-Fei. Im- agenet: A large-scale hierarchical image database. In CVPR,

work page

[7] [7]

H. Fan, X. Chang, D. Cheng, Y . Yang, D. Xu, and A. G. Hauptmann. Complex event detection by identifying reliable shots from untrimmed videos. In ICCV, 2017. 2

work page 2017

[8] [8]

H. Fan, L. Zheng, C. Yan, and Y . Yang. Unsupervised person re-identiﬁcation: Clustering and ﬁne-tuning. ACM Transac- tions on Multimedia Computing, Communications, and Ap- plications (TOMM), 14(4):83, 2018. 2

work page 2018