Pith. sign in

REVIEW 4 major objections 5 minor 62 references

Style-Adaptive Detection Transformer for Single-Source Domain Generalized Object Detection

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · 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 a DETR-based detector with an online style adapter and object-aware contrastive learning becomes the best single-source domain generalized object detector across five weather scenarios.

desk verdict A real DETR-based SDG method with code out, but the headline comparison is not apples-to-apples: the model uses test-time adaptation while baselines do not. read the letter →

arxiv 2504.20498 v2 pith:52UUXZYO submitted 2025-04-29 cs.CV

Jianhong Han , Yupei Wang , Liang Chen This is my paper · ORCID
classification cs.CV
keywords single-sourcedomaingeneralizationobjectdetectiontransformertest-timeadaptationstylecontrastivelearningweathershiftDINO
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 argues that for single-source domain generalization in object detection, the right strategy is not to simulate unseen domains through augmentation but to correct the style of whatever target appears, at test time, using statistics learned from the source. It proposes SA-DETR, a DETR-based detector that keeps a memory bank of channel-wise mean and variance prototypes of backbone features, projects unseen target features onto the source style manifold with a Wasserstein-weighted AdaIN operation, and updates the bank during inference so large style gaps are not under-adapted. An object-aware contrastive learning module is added during training to align instance-level features across domains. On a five-scenario weather benchmark trained only on daytime-clear images, the method reports the best mAP50 in every target, including an 8.4-point improvement on dusk-rainy detection. A sympathetic reader would care because this points to a deployment robustness recipe that needs no target-domain labels.

What carries the argument

The central mechanism is the Online Domain Style Adapter: a ring-buffer memory bank of $K=4$ style prototypes, each a channel-wise mean and variance vector computed from backbone feature maps. A sample's statistics are compared to the prototypes by Wasserstein distance; an adaptive threshold decides whether to fuse the sample into the nearest prototype by exponential moving average or to evict the least-used prototype and store the new statistics. At inference, softmax-normalized distances weight the prototypes, and the feature map is rectified as $F_{\text{rect}} = \frac{F - \mu}{\sigma} \sigma' + \mu'$, an AdaIN-style projection that maps the test image's style toward the source distribution. The supporting mechanism is the Object-aware Contrastive Learning module, in which class queries inserted into the encoder aggregate only object regions via gating masks built from bounding-box annotations, and a contrastive loss aligns same-class query prototypes between the source and augmented images.

What would settle it

Train SA-DETR on Daytime-Clear and test the ODS-Adapter on a genuinely novel weather not in the four targets, such as overnight snowfall, then compare adapted against unadapted mAP50 and track the memory bank's prototypes. If the bank's Wasserstein distance to source prototypes grows over the test stream while mAP50 does not improve or falls, the style-statistics premise and the fusion-only update rule would be refuted.

Watch

Extended reading notes

Core claim

SA-DETR claims that a DETR-based detector, specifically DINO, can be made the strongest single-source domain generalized object detector by decoupling the problem into a style-level correction and an instance-level alignment. The Online Domain Style Adapter treats the channel-wise mean and standard deviation of backbone feature maps as a style representation, stores them in a self-organizing memory bank of K style prototypes, and, at inference, remaps any test image's statistics to the source manifold via a weighted AdaIN projection whose weights come from softmax-normalized Wasserstein distances. Because entirely novel styles would otherwise under-adapt, the adapter also writes target statistics back into the bank under a test-time adaptation scheme, using only fusion so anomalous samples do not contaminate the prototypes. The object-aware contrastive learning module inserts per-class queries into the transformer encoder, restricts their aggregation with annotation-derived gating masks, and pulls same-category queries together while pushing different categories apart across the source and augmented domains. The paper reports best results in all five weather scenarios, with the largest gain on Dusk-Rainy where mAP50 rises to 46.5, an 8.4-point improvement over the previous best.

Load-bearing premise

The load-bearing premise is that channel-wise mean and variance of backbone features capture enough of a domain's style that re-matching them to a learned bank via a weighted AdaIN projection removes the domain gap, and that feeding target statistics back into the bank during test-time adaptation improves rather than pollutes the estimate.

Editorial extensions

If this is right

  • A detector trained only on clear daytime urban images can be deployed into fog, dusk-rain, and night conditions without retraining or target labels, and still improve over the unadapted DETR baseline in every tested scenario.
  • Because the style adapter operates on feature statistics rather than generated images, its inference overhead is small: 87.4 ms per image versus 81.2 ms for vanilla DINO, with test-time memory updates adding about 1 ms.
  • The object-aware contrastive loss is train-time only, so the deployed model keeps DINO's inference architecture plus the adapter; no extra module needs to run at inference.
  • The memory bank's self-organization matters: collapsing to a single style prototype, $K=1$, drops average target mAP50 from 44.52 to 42.72, and using more than four prototypes also hurts, indicating that the source domain supplies limited but real style diversity.
  • Style rectification is most valuable on hard low-light domains: Night-Rainy rises from 17.3 for vanilla DINO to 24.5, and Dusk-Rainy from 35.7 to 46.5, while the near-source Daytime-Foggy gain is smaller.

Reading between the lines

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

  • A direct testable extension: if channel-wise statistics truly span the style space, the same adapter should transfer to other visual shifts such as snow, sandstorms, or sensor artifacts without re-training, provided the new styles remain near the source manifold.
  • The fusion-only update in the test-time phase is a quietly load-bearing design choice: it assumes anomalous test images are rare, so a stress test with adversarial or heavily corrupted batches would show whether the bank drifts despite the fusion-only rule.
  • The OCL design suggests a general pattern for DETR-based generalization: use weakly supervised gating to build instance-level contrast, which could extend to segmentation or open-vocabulary detection where categories are not fixed.
  • The results leave open whether the gain comes mostly from the style adapter or from DINO's global attention; comparing a CNN detector equipped with the same ODS-Adapter would isolate the contribution of the transformer backbone.
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.

Referee Report

4 major / 5 minor

Summary. The paper proposes SA-DETR, a DETR-based detector for single-source domain generalization (SDG) in object detection. SA-DETR uses a DINO backbone and adds two components: an Online Domain Style Adapter (ODS-Adapter) that rectifies channel-wise feature statistics toward stored source style prototypes via a dynamic memory bank, with test-time updates of the bank; and an Object-aware Contrastive Learning (OCL) module that aligns instance-level features across source and augmented domains using class queries and gating masks. Experiments on a five-weather benchmark (Daytime-Clear as source; Daytime-Foggy, Dusk-Rainy, Night-Clear, Night-Rainy as targets) show consistent mAP improvements over prior SDG methods, with a claimed 8.4% mAP gain on Dusk-Rainy. The paper also reports ablations of the adapter, the number of style bases, and the temperature coefficient, plus inference-time overhead measurements.

Significance. If the central claims hold, the paper would be the first DETR-based detector for single-source domain generalized object detection and would demonstrate that test-time style-statistics adaptation plus contrastive instance alignment is a strong recipe for this benchmark. The work has several strengths: the method is clearly motivated, the code is released, the ablations in Table I are informative, and the source-domain performance is not sacrificed. However, the evaluation protocol weakens the SDG claim: the memory bank is updated on target test data at inference (a test-time adaptation, TTA, protocol), and two key hyperparameters (K and alpha) are selected by scanning mAP50 on the four unseen target domains. The headline improvements therefore conflate learned generalization with online adaptation to the test stream and with test-set fitting.

major comments (4)
  1. [Section III-B, Eq. (3)-(8), and Tables I-VII] The evaluation is not an apples-to-apples SDG comparison. The manuscript states in Section III-B that "During testing, we employ the same update strategy to enable the model to rapidly and continuously absorb the style characteristics of unseen scenes," and Table I shows that enabling TTA raises Night-Rainy mAP from 20.7 to 22.9. The baseline SDG methods (SDGOD, UFR, CLIP-Gap, etc.) are trained on source data and deployed without any target information, so the reported gains, including the headline 8.4% improvement on Dusk-Rainy, are not attributable solely to domain-generalizable learned features. The authors should report a "no-TTA" version of SA-DETR (memory bank fixed after training) with the identical architecture, compare against TTA-based object detection methods, or explicitly reframe the claim as a TTA-inclusive evaluation. Without this, the statement that the method "consistently achieved the best outcomes in all scenarios" is not a valid SDG result.
  2. [Tables II and III] The hyperparameters K (number of style bases) and alpha (temperature coefficient) are selected by scanning mAP50 on the four unseen target domains. Table II chooses K=4 because it yields the "peak average" of 44.52%, and Table III chooses alpha=0.7 because it gives the "best average mAP50" of 44.52%. This is a form of test-set fitting: the model is tuned on the very domains it is supposed to generalize to. To support the claim of generalization to unseen domains, K and alpha should be fixed using a validation split from the source domain, or justified as constants that are not dataset-dependent. If test-set selection was unavoidable, the paper must disclose this and show sensitivity over a range rather than presenting the selected values as if they were model hyperparameters.
  3. [Section III-B, paragraphs on the memory bank and TTA] There is an internal inconsistency in the description of the test-time update rule. The paragraph says "During testing, we employ the same update strategy" (same as training, which includes replacing the least-frequently-used prototype when d_min > tau), but then immediately states "It is noteworthy that only the fusion is employed to mitigate the contamination of the memory bank caused by discrete or anomalous samples during the TTA phase." These two sentences describe incompatible behaviors. Since the test-time behavior is load-bearing for the night-scene results, the authors must specify the exact TTA algorithm (e.g., a pseudocode block) and clarify whether replacement is disabled, and if so, how the bank "continuously absorbs" genuinely novel styles.
  4. [Section III-C, Eq. (11)] The contrastive loss in Eq. (11) is not fully specified. The formulation sums over all C categories in the denominator, but the text says "we compute the loss only for the categories that are present in the images." If a category is absent from an image, its class query has no valid aggregated feature (the gating mask is all zeros), so it is unclear what q^S_j or q^A_i denote in the denominator for those categories, and whether the loss is computed per-image or per-batch. This needs clarification, ideally with an explicit index set of present categories in both numerator and denominator, otherwise the OCL module's mechanism is ambiguous.
minor comments (5)
  1. [Eq. (1) and Eq. (2)] The summations in Eq. (1) and Eq. (2) write F without spatial indices; they should be F_{h,w} or F[:, :, h, w] to be mathematically precise.
  2. [Reference [53]] Reference [53] is cited for exponential moving average, but Marsella and Gratch's EMA is a process model of appraisal dynamics in cognitive science, not an exponential moving average. A standard reference for EMA, such as the momentum update in [52], should be used.
  3. [Table I and Section IV.C.1] The abbreviations "Day-F", "Dusk-R", "Night-R", and "Night-C" are used without introduction; they should be defined in the table caption or in the text.
  4. [Section III-B, soft-KNN sentence] The sentence "Additionally, we explore a soft-KNN retrieval mechanism that selects a small set of representative bases instead of utilizing the entire bank, which is found to be suboptimal" is not supported by any ablation or table. Either add the corresponding experiment or remove the claim.
  5. [General] No standard deviations or multiple-seed results are reported. Given the small margins on some target domains (e.g., 0.4% on Night-Rainy in Table VII), the claim of consistent superiority would be strengthened by reporting mean and variance over at least three random seeds.

Circularity Check

1 steps flagged · score 6.0 of 10

The headline benchmark results are partly fitted to the target test sets: K and alpha are selected using mAP50 on the four 'unseen' domains, then reported as generalization; the TTA memory update additionally makes the SDG comparison non-apples-to-apples.

  1. fitted input called prediction [Section IV-C, Tables II-III; final comparisons in Tables V-IX]
    "When k = 1 , all style statistics are collapsed into a single prototype, similar to existing methods [9], [10], [40], resulting in a sub-optimal average mAP50 of 42.72%. Increasingk from 1 to 4 steadily improves performance, with the largest gain appearing on the most challenging domains and a peak average of 44.52% at k = 4. ... The best averagemAP50 is achieved when α = 0 .7, largely due to more substantial gains on challenging low-light domains such as Dusk-R and Night-R."

    The paper chooses K=4 and alpha=0.7 by scanning the average mAP50 over the four target test domains and picking the maximum (44.52% in both Tables II and III). Those same target domains are then used in Tables V-IX to report SA-DETR's 'superior performance and generalization capability', including the 8.4% Dusk-Rainy gain. The reported 'prediction' is therefore not an out-of-sample result: the two hyperparameters of the final model were optimized on the exact benchmark that the paper claims to predict. This is fitted input presented as prediction, not a derivation from first principles.

full rationale

The method's equations are not self-referential: the AdaIN projection (Eqs. 7-8) is ordinary statistical modulation driven by a memory bank, and the contrastive objective (Eq. 11) is a standard auxiliary loss. No central claim is forced by a self-citation chain; the only self-citation, [18], supports a generic statement about transformer global modeling and is not load-bearing. However, the empirical evaluation is partially circular in the fitted-input sense: K and alpha are tuned on the four target test domains (Tables II-III) and the tuned values are reused in the headline comparisons (Tables V-IX). A separate, non-circular comparability problem is that the ODS-Adapter performs test-time adaptation by writing target statistics into the memory bank ('During testing, we employ the same update strategy...'), so the SDG comparison mixes source-trained generalization with online adaptation to the test stream. Because the architecture retains independent content but the headline numbers are partly bench-fitted, the circularity score is 6 rather than higher.

Assumptions & free parameters 4 free parameters · 5 assumptions · 3 invented entities

The method relies on a compact statistical-style assumption, a test-time update assumption, and four hyperparameters, two of which were tuned directly on the target test sets. The invented entities are internal algorithmic components with no independent evidence outside the evaluated pipeline.

free parameters (4)
  • Memory bank size K = 4 (selected by target-domain mAP50, Table II)
    Grid search over K=1..8 on the four unseen target test sets; K=4 gives the best average mAP50 of 44.52%, and all subsequent experiments use K=4.
  • Temperature coefficient alpha = 0.7 (selected by target-domain mAP50, Table III)
    Controls the adaptive threshold in Eq. 4; alpha=0.7 gives the best average mAP50 of 44.52%, so it is tuned on the test domains.
  • EMA momentum lambda = not stated in paper
    Used in Eqs. 5-6 for fusing statistics into the memory bank. Its value is not reported, so the behavior of the bank depends on an undisclosed hyperparameter.
  • Contrastive loss weight lambda_c = 0.1
    Balancing weight in Eq. 12; set to 0.1 for all scenarios with no ablation study, so its contribution to the reported performance is not independently characterized.
assumptions (5)
  • domain assumption Channel-wise mean and variance of backbone features are a sufficient representation of domain style.
    The entire style adapter is built on computing mean and variance in Eqs. 1-2 and using them as style statistics in the AdaIN projection of Eqs. 7-8.
  • domain assumption Wasserstein distance between channel-wise statistics is a valid measure of style discrepancy.
    Used in Eq. 3 to compare input statistics with stored prototypes, to decide memory-bank updates, and to compute AdaIN weights.
  • domain assumption The spurious-correlation augmentation from MAD [32] creates a meaningful 'augmented domain' for contrastive learning.
    The object-aware contrastive loss in Eq. 11 aligns features between source and augmented domains; if this augmentation is not a useful proxy for unseen domains, the OCL module loses its purpose.
  • domain assumption Updating the memory bank with target statistics during test time improves adaptation without contaminating the bank.
    Section III-B states that during testing the same update strategy is used, with only fusion employed to mitigate contamination; the stability of this online update is not proven.
  • domain assumption Transformer global attention in DINO provides better domain generalization than CNN detectors.
    The paper motivates the method by citing DETR's strong generalization in domain adaptation (Section II-A) and uses DINO as the base, but no new evidence is supplied beyond the empirical comparisons.
invented entities (3)
  • Dynamic style memory bank of style prototypes
    purpose: Stores channel-wise mean/variance statistics as style bases and serves as the reference manifold for weighted AdaIN projection.
    The bank is internal to the method; its only support is the reported mAP results, and it has no external falsifiable handle outside the proposed system.
  • Learnable class queries inserted into the encoder
    purpose: Aggregate instance-level features for each category via cross-attention with gating masks, enabling cross-domain contrastive learning.
    These are new latent vectors introduced for the OCL module; their behavior is only measured indirectly through downstream detection accuracy.
  • Object-aware gating masks
    purpose: Restrict feature aggregation to annotated object locations and categories in both spatial and semantic dimensions.
    The masks are constructed from training annotations and used inside the module; they are an algorithmic device rather than an independently verified entity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Style-Adaptive Detection Transformer for Single-Source Domain Generalized Object Detection." pith.science (2026). https://pith.science/paper/52UUXZYO

@misc{pith2026250420498,
  author       = {Pith},
  title        = {Pith review of: Style-Adaptive Detection Transformer for Single-Source Domain Generalized Object Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/52UUXZYO}},
  note         = {Machine review of arXiv:2504.20498}
}
read the original abstract

Single-source domain generalization (SDG) in object detection aims to develop a detector using only source domain data that generalizes well to unseen target domains. Existing methods are primarily CNN-based and improve robustness through data augmentation combined with feature alignment. However, these methods are limited, as augmentation is only effective when the synthetic distribution approximates that of unseen domains, thus failing to ensure generalization across diverse scenarios. While DEtection TRansformer (DETR) has shown strong generalization in domain adaptation due to global context modeling, its potential for SDG remains underexplored. To this end, we propose Style-Adaptive DEtection TRansformer (SA-DETR), a DETR-based detector tailored for SDG. SA-DETR introduces an online domain style adapter that projects the style representation of unseen domains into the source domain via a dynamic memory bank. This bank self-organizes into diverse style prototypes and is continuously updated under a test-time adaptation framework, enabling effective style rectification. Additionally, we design an object-aware contrastive learning module to promote extraction of domain-invariant features. By applying gating masks that constrain contrastive learning in both spatial and semantic dimensions, this module facilitates instance-level cross-domain contrast and enhances generalization. Extensive experiments across five distinct weather scenarios demonstrate that SA-DETR consistently outperforms existing methods in both detection accuracy and domain generalization capability.

Figures

Figures reproduced from arXiv: 2504.20498 by the authors.

Figure 1
Figure 1. Details of (a) the overview of the proposed SA-DETR framework, which adopts DINO as the base detector and [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Our proposed object-aware contrastive learning (OCL) module. Details of (a) the inserted class queries adaptively [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 4
Figure 4. The t-SNE visualization of object features from images [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figures from the paper (3 more)
Figure 5
Figure 5. Figure 5: The t-SNE visualization of object features from the five [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 3
Figure 3. Figure 3: The t-SNE visualization of style statistics between [PITH_FULL_IMAGE:figures/full_fig_p010_3.png]
Figure 6
Figure 6. Figure 6: Our detection results compared with the state-of-the-art methods. Different categories are marked with different colors. [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

62 extracted references · 37 canonical work pages

  1. [1]

    Domain adaptive faster r-cnn for object detection in the wild,

    Y . Chen, W. Li, C. Sakaridis, D. Dai, and L. Van Gool, “Domain adaptive faster r-cnn for object detection in the wild,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2018, pp. 3339–3348

    2018

  2. [2]

    Da- detr: Domain adaptive detection transformer with information fusion,

    J. Zhang, J. Huang, Z. Luo, G. Zhang, X. Zhang, and S. Lu, “Da- detr: Domain adaptive detection transformer with information fusion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 23 787–23 798

    2023

  3. [3]

    Mttrans: Cross-domain object detection with mean teacher transformer,

    J. Yu, J. Liu, X. Wei, H. Zhou, Y . Nakata, D. Gudovskiy, T. Okuno, J. Li, K. Keutzer, and S. Zhang, “Mttrans: Cross-domain object detection with mean teacher transformer,” in Proceedings of the European Conference on Computer Vision , 2022, pp. 629–645

    work page 2022

  4. [4]

    Single-domain generalized object detection in urban scene via cyclic-disentangled self-distillation,

    A. Wu and C. Deng, “Single-domain generalized object detection in urban scene via cyclic-disentangled self-distillation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 847–856

    work page 2022

  5. [5]

    Clip the gap: A single domain generalization approach for object detection,

    V . Vidit, M. Engilberge, and M. Salzmann, “Clip the gap: A single domain generalization approach for object detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 3219–3229

    work page 2023

  6. [6]

    Unbiased faster r-cnn for single-source domain generalized object detection,

    Y . Liu, S. Zhou, X. Liu, C. Hao, B. Fan, and J. Tian, “Unbiased faster r-cnn for single-source domain generalized object detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 28 838–28 847

    2024

  7. [7]

    Two at once: Enhancing learning and generalization capacities via ibn-net,

    X. Pan, P. Luo, J. Shi, and X. Tang, “Two at once: Enhancing learning and generalization capacities via ibn-net,” in Proceedings of the Euro- pean Conference on Computer Vision , 2018, pp. 464–479

    work page 2018

  8. [8]

    Robustnet: Improving domain generalization in urban-scene segmentation via in- stance selective whitening,

    S. Choi, S. Jung, H. Yun, J. T. Kim, S. Kim, and J. Choo, “Robustnet: Improving domain generalization in urban-scene segmentation via in- stance selective whitening,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 11 580–11 590

    work page 2021

Show all 62 references
  1. [9]

    Fsda-detr: Few- shot domain-adaptive object detection transformer in remote sensing im- agery,

    B. Yang, J. Han, X. Hou, D. Zhou, W. Liu, and F. Bi, “Fsda-detr: Few- shot domain-adaptive object detection transformer in remote sensing im- agery,” IEEE Transactions on Geoscience and Remote Sensing , vol. 63, pp. 1–16, 2025

    2025

  2. [10]

    Losa: Learnable online 12 style adaptation for test-time domain adaptive object detection,

    W. Liu, B. Luo, J. Liu, H. Nie, and X. Su, “Losa: Learnable online 12 style adaptation for test-time domain adaptive object detection,” IEEE Transactions on Geoscience and Remote Sensing , vol. 63, pp. 1–16, 2025

    2025

  3. [11]

    Faster r-cnn: Towards real-time object detection with region proposal networks,

    S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real-time object detection with region proposal networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 39, no. 6, pp. 1137– 1149, 2017

    2017

  4. [12]

    Exploring sequence feature alignment for domain adaptive detection transformers,

    W. Wang, Y . Cao, J. Zhang, F. He, Z.-J. Zha, Y . Wen, and D. Tao, “Exploring sequence feature alignment for domain adaptive detection transformers,” in Proceedings of the ACM International Conference on Multimedia, 2021, pp. 1730–1738

    2021

  5. [13]

    Aqt: Adversarial query transformers for domain adaptive object detection,

    W.-J. Huang, Y .-L. Lu, S.-Y . Lin, Y . Xie, and Y .-Y . Lin, “Aqt: Adversarial query transformers for domain adaptive object detection,” inProceedings of the International Joint Conference on Artificial Intelligence, 2022, pp. 972–979

    2022

  6. [14]

    Masked retraining teacher-student framework for domain adaptive object detec- tion,

    Z. Zhao, S. Wei, Q. Chen, D. Li, Y . Yang, Y . Peng, and Y . Liu, “Masked retraining teacher-student framework for domain adaptive object detec- tion,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 19 039–19 049

    2023

  7. [15]

    End-to-end object detection with transformers,

    N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, and S. Zagoruyko, “End-to-end object detection with transformers,” in Proceedings of the European Conference on Computer Vision. Springer, 2020, pp. 213–229

    2020

  8. [16]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Transformers for image recognition at scale,” in Proceedings of the International Con...

    2021

  9. [17]

    Aloft: A lightweight mlp-like architecture with dynamic low-frequency transform for domain gener- alization,

    J. Guo, N. Wang, L. Qi, and Y . Shi, “Aloft: A lightweight mlp-like architecture with dynamic low-frequency transform for domain gener- alization,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 24 132–24 141

    2023

  10. [18]

    Encouraging the mutual interact between dataset-level and image-level context for semantic segmentation of remote sensing image,

    K. An, Y . Wang, and L. Chen, “Encouraging the mutual interact between dataset-level and image-level context for semantic segmentation of remote sensing image,” IEEE Transactions on Geoscience and Remote Sensing, vol. 62, no. 5606116, pp. 1–16, 2024

    2024

  11. [19]

    Domain-aware prototype network for generalized zero-shot learning,

    Y . Hu, L. Feng, H. Jiang, M. Liu, and B. Yin, “Domain-aware prototype network for generalized zero-shot learning,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 34, no. 5, pp. 3180– 3191, 2024

    2024

  12. [20]

    Arbitrary style transfer in real-time with adaptive instance normalization,

    X. Huang and S. Belongie, “Arbitrary style transfer in real-time with adaptive instance normalization,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2017, pp. 1501–1510

    2017

  13. [21]

    Banach wasserstein gan,

    J. Adler and S. Lunz, “Banach wasserstein gan,” Advances in neural information processing systems , vol. 31, 2018

    2018

  14. [22]

    Uncertainty-aware unsupervised domain adaptation in object detection,

    D. Guan, J. Huang, A. Xiao, S. Lu, and Y . Cao, “Uncertainty-aware unsupervised domain adaptation in object detection,” IEEE Transactions on Multimedia, vol. 24, pp. 2502–2514, 2021

    2021

  15. [23]

    Self-labeling framework for open-set domain adaptation with few labeled samples,

    Q. Yu, G. Irie, and K. Aizawa, “Self-labeling framework for open-set domain adaptation with few labeled samples,” IEEE Transactions on Multimedia, vol. 26, pp. 1474–1487, 2023

    2023

  16. [24]

    Madav2: Advanced multi-anchor based active domain adaptation segmentation,

    M. Ning, D. Lu, Y . Xie, D. Chen, D. Wei, Y . Zheng, Y . Tian, S. Yan, and L. Yuan, “Madav2: Advanced multi-anchor based active domain adaptation segmentation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 11, pp. 13 553–13 566, 2023

    2023

  17. [25]

    Adversarially adaptive normalization for single domain generalization,

    X. Fan, Q. Wang, J. Ke, F. Yang, B. Gong, and M. Zhou, “Adversarially adaptive normalization for single domain generalization,” inProceedings of the IEEE/CVF conference on Computer Vision and Pattern Recogni- tion, 2021, pp. 8208–8217

    2021

  18. [26]

    Exact feature distribution matching for arbitrary style transfer and domain generalization,

    Y . Zhang, M. Li, R. Li, K. Jia, and L. Zhang, “Exact feature distribution matching for arbitrary style transfer and domain generalization,” in Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition, 2022, pp. 8035–8045

    2022

  19. [27]

    Progressive diversity generation for single domain generalization,

    D. Rui, K. Guo, X. Zhu, Z. Wu, and H. Fang, “Progressive diversity generation for single domain generalization,” IEEE Transactions on Multimedia, pp. 1–12, 2024

    2024

  20. [28]

    Adversarial source generation for source-free domain adaptation,

    C. Cui, F. Meng, C. Zhang, Z. Liu, L. Zhu, S. Gong, and X. Lin, “Adversarial source generation for source-free domain adaptation,”IEEE Transactions on Circuits and Systems for Video Technology , vol. 34, no. 6, pp. 4887–4898, 2024

    2024

  21. [29]

    Style projected clustering for domain generalized semantic segmenta- tion,

    W. Huang, C. Chen, Y . Li, J. Li, C. Li, F. Song, Y . Yan, and Z. Xiong, “Style projected clustering for domain generalized semantic segmenta- tion,” in Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition, 2023, pp. 3061–3071

    2023

  22. [30]

    Wildnet: Learning domain generalized semantic segmentation from the wild,

    S. Lee, H. Seong, S. Lee, and E. Kim, “Wildnet: Learning domain generalized semantic segmentation from the wild,” in Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition , 2022, pp. 9936–9946

    2022

  23. [31]

    Learning generalized knowledge from a single domain on urban-scene segmentation,

    X. Li, M. Li, X. Li, and X. Guo, “Learning generalized knowledge from a single domain on urban-scene segmentation,” IEEE Transactions on Multimedia, vol. 25, pp. 7635–7646, 2023

    2023

  24. [32]

    Multi-view adversarial discriminator: Mine the non-causal factors for object detection in unseen domains,

    M. Xu, L. Qin, W. Chen, S. Pu, and L. Zhang, “Multi-view adversarial discriminator: Mine the non-causal factors for object detection in unseen domains,” in Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition , 2023, pp. 8103–8112

    2023

  25. [33]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Loet al., “Segment anything,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 4015–4026

    2023

  26. [34]

    Source-free domain adaptive object detection in remote sensing images,

    W. Liu, J. Liu, X. Su, H. Nie, and B. Luo, “Source-free domain adaptive object detection in remote sensing images,” arXiv preprint arXiv:2401.17916, 2024

    2024 arXiv

  27. [35]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark et al., “Learning transferable visual models from natural language supervision,” in Proceedings of the International conference on machine learning , 2021, pp. 8748–8763

    2021

  28. [36]

    Phrase grounding-based style transfer for single-domain generalized object detection,

    H. Li, W. Wang, C. Wang, Z. Luo, X. Liu, K. Li, and X. Cao, “Phrase grounding-based style transfer for single-domain generalized object detection,” arXiv preprint arXiv:2402.01304 , 2024

    2024

  29. [37]

    Grounded language-image pre- training,

    L. H. Li, P. Zhang, H. Zhang, J. Yang, C. Li, Y . Zhong, L. Wang, L. Yuan, L. Zhang, J.-N. Hwang et al., “Grounded language-image pre- training,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 10 965–10 975

    2022

  30. [38]

    Domain generalization with mixstyle,

    K. Zhou, Y . Yang, Y . Qiao, and T. Xiang, “Domain generalization with mixstyle,” arXiv preprint arXiv:2104.02008 , 2021

    2021 arXiv

  31. [39]

    Learning frequency-adapted vision foundation model for domain generalized semantic segmentation,

    Q. Bi, J. Yi, H. Zheng, H. Zhan, Y . Huang, W. Ji, Y . Li, and Y . Zheng, “Learning frequency-adapted vision foundation model for domain generalized semantic segmentation,” Advances in Neural Infor- mation Processing Systems , vol. 37, pp. 94 047–94 072, 2024

    2024

  32. [40]

    Domain-rectifying adapter for cross-domain few-shot segmentation,

    J. Su, Q. Fan, W. Pei, G. Lu, and F. Chen, “Domain-rectifying adapter for cross-domain few-shot segmentation,” in Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition , 2024, pp. 24 036–24 045

    2024

  33. [41]

    Domain adaptation for object detection via style consistency,

    A. L. Rodriguez and K. Mikolajczyk, “Domain adaptation for object detection via style consistency,” arXiv preprint arXiv:1911.10033, 2019

    1911 arXiv

  34. [42]

    Cross domain object detection by target-perceived dual branch distillation,

    M. He, Y . Wang, J. Wu, Y . Wang, H. Li, B. Li, W. Gan, W. Wu, and Y . Qiao, “Cross domain object detection by target-perceived dual branch distillation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 9570–9580

    2022

  35. [43]

    Tent: Fully test-time adaptation by entropy minimization,

    D. Wang, E. Shelhamer, S. Liu, B. Olshausen, and T. Darrell, “Tent: Fully test-time adaptation by entropy minimization,” arXiv preprint arXiv:2006.10726, 2020

    2006 arXiv

  36. [44]

    The norm must go on: Dynamic unsupervised domain adaptation by normalization,

    M. J. Mirza, J. Micorek, H. Possegger, and H. Bischof, “The norm must go on: Dynamic unsupervised domain adaptation by normalization,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 14 765–14 775

    2022

  37. [45]

    Test-time adaptation for real-world video adverse weather restoration with meta batch normalization,

    J. Liu and Z. Yang, “Test-time adaptation for real-world video adverse weather restoration with meta batch normalization,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 35, no. 6, pp. 5533– 5544, 2025

    2025

  38. [46]

    Actmad: Activation matching to align distributions for test-time-training,

    M. J. Mirza, P. J. Soneira, W. Lin, M. Kozinski, H. Possegger, and H. Bischof, “Actmad: Activation matching to align distributions for test-time-training,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 24 152–24 161

    2023

  39. [47]

    Towards online domain adaptive object detection,

    V . VS, P. Oza, and V . M. Patel, “Towards online domain adaptive object detection,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , 2023, pp. 478–488

    2023

  40. [48]

    Exploring test-time adaptation for object detection in continually changing environments,

    S. Cao, Y . Liu, J. Zheng, W. Li, R. Dong, and H. Fu, “Exploring test-time adaptation for object detection in continually changing environments,” arXiv preprint arXiv:2406.16439 , 2024

    2024 arXiv

  41. [49]

    Mlfa: Toward realistic test time adaptive object detection by multi-level feature alignment,

    Y . Liu, J. Wang, C. Huang, Y . Wu, Y . Xu, and X. Cao, “Mlfa: Toward realistic test time adaptive object detection by multi-level feature alignment,” IEEE Transactions on Image Processing, vol. 33, pp. 5837– 5848, 2024

    2024

  42. [50]

    Dino: Detr with improved denoising anchor boxes for end-to- end object detection,

    H. Zhang, F. Li, S. Liu, L. Zhang, H. Su, J. Zhu, L. M. Ni, and H.-Y . Shum, “Dino: Detr with improved denoising anchor boxes for end-to- end object detection,” in Proceedings of the International Conference on Learning Representations , 2023

    2023

  43. [51]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition , 2016, pp. 770–778

    2016

  44. [52]

    Momentum contrast for unsupervised visual representation learning,

    K. He, H. Fan, Y . Wu, S. Xie, and R. Girshick, “Momentum contrast for unsupervised visual representation learning,” in 2020 IEEE/CVF 13 Conference on Computer Vision and Pattern Recognition , 2020, pp. 9726–9735

    2020

  45. [53]

    Ema: A process model of appraisal dynamics,

    S. C. Marsella and J. Gratch, “Ema: A process model of appraisal dynamics,” Cognitive Systems Research , vol. 10, no. 1, p. 70–90, Mar 2009

    2009

  46. [54]

    Bdd100k: A diverse driving dataset for heterogeneous multitask learning,

    F. Yu, H. Chen, X. Wang, W. Xian, Y . Chen, F. Liu, V . Madhavan, and T. Darrell, “Bdd100k: A diverse driving dataset for heterogeneous multitask learning,” in Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition , 2020, pp. 2636–2645

    2020

  47. [55]

    The cityscapes dataset for semantic urban scene understanding,

    M. Cordts, M. Omran, S. Ramos, T. Rehfeld, M. Enzweiler, R. Benen- son, U. Franke, S. Roth, and B. Schiele, “The cityscapes dataset for semantic urban scene understanding,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2016, pp. 3213–3223

    2016

  48. [56]

    Vehi- cle detection and tracking in adverse weather using a deep learning framework,

    M. Hassaballah, M. A. Kenk, K. Muhammad, and S. Minaee, “Vehi- cle detection and tracking in adverse weather using a deep learning framework,” IEEE Transactions on Intelligent Transportation Systems , p. 4230–4242, 2021

    2021

  49. [57]

    Adam: A method for stochastic optimization,

    D. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

  50. [58]

    Switchable whitening for deep representation learning,

    X. Pan, X. Zhan, J. Shi, X. Tang, and P. Luo, “Switchable whitening for deep representation learning,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2019, pp. 1863–1871

    2019

  51. [59]

    Iterative normalization: Beyond standardization towards efficient whitening,

    L. Huang, Y . Zhou, F. Zhu, L. Liu, and L. Shao, “Iterative normalization: Beyond standardization towards efficient whitening,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 4874–4883

    2019

  52. [60]

    Improving single domain-generalized object detection: A focus on diversification and alignment,

    M. S. Danish, M. H. Khan, M. A. Munir, M. S. Sarfraz, and M. Ali, “Improving single domain-generalized object detection: A focus on diversification and alignment,” in 2024 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , 2024, pp. 17 732–17 742

    2024

  53. [61]

    Srcd: Se- mantic reasoning with compound domains for single-domain generalized object detection,

    Z. Rao, J. Guo, L. Tang, Y . Huang, X. Ding, and S. Guo, “Srcd: Se- mantic reasoning with compound domains for single-domain generalized object detection,” IEEE Transactions on Neural Networks and Learning Systems, vol. 36, no. 7, pp. 12 497–12 506, 2025

    2025

  54. [62]

    Visualizing data using t-sne

    L. Van der Maaten and G. Hinton, “Visualizing data using t-sne.” Journal of machine learning research , vol. 9, no. 11, 2008

    2008

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.