Pith. sign in

REVIEW 3 major objections 5 minor 68 references

Language-Driven Dual Style Mixing for Single-Domain Generalized Object Detection

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Language prompts plus dual style mixing let object detectors generalize to unseen domains without changing their backbone.

desk verdict A genuinely useful model-agnostic framework for VLM-based single-domain generalized detection, but the headline gains rest partly on hyperparameters tuned on a test set, so the empirical claims need a clean rerun before they're credible. read the letter →

arxiv 2505.07219 v1 pith:QHBM65L3 submitted 2025-05-12 cs.CV cs.ROeess.IV

Hongda Qin , Xiao Lu , Zhiyong Wei , Yihong Cao , Kailun Yang , Ningjiang Chen This is my paper · ORCID
classification cs.CVcs.ROeess.IV
keywords single-domaingeneralizationobjectdetectionvision-languagemodelCLIPstylemixingFourieramplitudefeaturestatisticsadverseweather
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 a detector trained on a single domain can generalize to unseen domains by mixing styles at both image and feature levels, with the style drawn from a vision-language model. The core claim is that CLIP's semantic understanding of a short text prompt such as "a photo of the road in the rain" can be transferred into training images, and that two rounds of mixing—one in the Fourier amplitude of images, one in the channel statistics of feature maps—let the detector absorb that style while keeping object details intact. If true, single-domain generalization no longer depends on the detector sharing an architecture with a vision-language encoder, so any mainstream detector can be upgraded without structural changes.

What carries the argument

The central object is the dual style mixing pair: image-level Fourier amplitude mixing and feature-level statistics mixing. The load-bearing identity is the Fourier decomposition in which phase retains content and amplitude carries global style, together with the observation that feature-map channel means and variances encode local style. The GMM smoothing applied to feature statistics of the style-diversified image is what prevents the second mixing round from re-introducing global style already mixed at the image level, thereby avoiding double-mixing conflicts.

What would settle it

Take a target domain that cannot be captured by a short English prompt, for instance a procedurally generated synthetic scene with a novel combination of lighting and texture. Run LDDS with the best available prompt; if the detection gain over the source-only baseline is no better than using a random or mismatched prompt, then the semantic content of the prompt is not what drives the improvement. A simpler check: measure object detection AP on the style-transferred images themselves; if StyleNet outputs already lose most objects, the mixing is preserving objects only because the Fourier amplitude mixing discards most of the style.

Watch

Extended reading notes

Core claim

LDDS claims to achieve strong single-domain generalized object detection by using language prompts to drive style generation, then mixing style at two levels. The image-level mixing blends the Fourier amplitude of the source image with that of a CLIP-guided style-transferred image, keeping the source phase to preserve content. The feature-level mixing runs a second pipeline for the style-transferred image and mixes its feature statistics (channel-wise mean and standard deviation) into the first pipeline's features, after a Gaussian Mixture Model smoothing step that removes redundant global style so the two mixing rounds do not conflict. The same recipe works for one-stage (YOLOv8), two-stage (Faster R-CNN), and transformer-based (RT-DETR) detectors, and it reports gains over prior methods on real-to-cartoon and normal-to-adverse-weather benchmarks.

Load-bearing premise

The whole gain rests on the assumption that an unseen target domain can be named ahead of time with a short phrase like "a photo of the road in the rain," and that CLIP-guided style transfer turns that phrase into images that carry the style without blurring or distorting the objects the detector must find.

Editorial extensions

If this is right

  • VLM-based semantic augmentation no longer binds the detector's backbone to the VLM's image encoder, so one-stage, two-stage, and transformer detectors can all benefit.
  • The night-rainy subset, the hardest among the adverse-weather splits, shows the largest relative gains, suggesting the method is most useful when the unseen domain is visually extreme.
  • Real-to-cartoon transfer improves across all three cartoon datasets for all three detector families, with the largest gains on Comic2k for RT-DETR.
  • Removing the GMM smoothing degrades performance, indicating that double-round style mixing without conflict handling is a real failure mode that the smoothing step addresses.

Reading between the lines

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

  • If the semantic content of the prompt is what carries the gain, then the method's ceiling depends on CLIP's ability to name a style; an unseen domain that resists a short verbal description, such as a novel synthetic or artistic genre, would likely see little benefit.
  • The same dual-mixing recipe could be ported to other single-domain generalization tasks like semantic segmentation or depth estimation, where feature statistics already play a central role.
  • Because the image-level mixing only swaps amplitudes, it is a cheap proxy for full style transfer; replacing StyleNet with a diffusion-based generator could push detail preservation further but at a much higher training-time cost.
  • The double-pipeline design requires generating a style-transferred image for each source image per batch, so training cost grows with the number of style prompts; batching multiple prompts per source image is a testable extension.
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

3 major / 5 minor

Summary. The manuscript proposes Language-Driven Dual Style Mixing (LDDS), a single-domain generalized object detection method that uses CLIP text prompts to drive a StyleNet to generate style-diversified images, then mixes styles at both image level (Fourier amplitude) and feature level (statistics mixing with GMM smoothing) to augment a source domain. Experiments cover real-to-cartoon (PASCAL VOC to Clipart1k/Watercolor2k/Comic2k) and normal-to-adverse-weather (BDD100k day-sunny to night-clear/dusk-rainy/night-rainy/daytime-foggy) with one-stage (YOLOv8), two-stage (Faster R-CNN), and transformer-based (RT-DETR) detectors. The paper claims state-of-the-art performance and model-agnostic applicability, and states that source code and models will be released.

Significance. The core idea is well motivated: removing the backbone-architecture constraint of prior VLM-based SDG detectors and using CLIP's semantic signal for both image and feature augmentation is a reasonable and potentially useful direction. The paper includes systematic ablations, visualizations, and results across three detector families, and if the findings hold it would be a valuable contribution to SDG object detection. That said, the empirical evidence is weakened by the fact that the mixing hyperparameters are selected using the night-rainy test set (Supplementary Sec. 6, Table 6) and then fixed for all reported experiments, as well as by the absence of error bars and the non-standard RT-DETR configuration. These issues must be addressed before the state-of-the-art claim can be accepted.

major comments (3)
  1. [Supplementary Sec. 6, Table 6] The sensitivity analysis in Table 6 is performed directly on the night-rainy test subset of BDD100k, and the chosen hyperparameters (gamma1=0.5, gamma2=1.0, beta1=0.1, beta2=2.0) are then used in all main experiments (Sec. 4.2). Since night-rainy is one of the unseen target domains evaluated in Tables 2, 3, and 7, the substantial night-rainy improvements (e.g., YOLOv8 from 13.7 to 19.1 in Table 3) are at least partly the result of target-domain test-set tuning, which violates the single-domain generalization protocol and inflates the headline gains on that domain. Please re-run the sensitivity analysis on a held-out validation set that is disjoint from all test domains (e.g., a split of the daytime-sunny source or an additional weather condition), report the parameters selected without any access to unseen-domain labels, and provide error bars over multiple runs. Without this, the state-of-the-art claim in the abstract and Section 4.3 is not fully supported.
  2. [Supplementary Sec. 7.1] The transformer-based experiments on RT-DETR are performed after replacing the original CNN-based HGNetv2 backbone with Swin Transformer-B. This is a non-trivial architectural change, and the baseline 'RT-DETR [61]' cited in Tables 1-3 presumably uses the standard HGNetv2 backbone. Consequently, the claim that LDDS works with a 'transformer-based detector' is only demonstrated on a modified RT-DETR configuration. Please clarify why this substitution was required, provide results on the standard RT-DETR architecture with HGNetv2, or at minimum clearly disclose in the main text that the RT-DETR experiments use a substituted backbone.
  3. [Sec. 4.3, Tables 1-2] Several reported gains are small (e.g., Table 2, Faster R-CNN on night-clear: 44.1 vs 42.5 for Div; YOLOv8 on daytime-sunny: 57.5 vs 57.3 baseline), yet the paper reports a single run with no error bars or statistical significance analysis. Combined with the test-set-based hyperparameter selection, this makes it difficult to assess which of the performance improvements are robust. Please include multiple seeds with standard deviations, or at least specify the number of runs, for the main results.
minor comments (5)
  1. [Sec. 4.3] The sentence 'compared to the recent methods, NP [17], Div [11], our LDDS achieves a gain of 14.4%, 14.9%, and 15.4% on Faster R-CNN' is ambiguous: 14.4%, 14.9%, and 15.4% appear to be the improvements over the source-only FRCNN baseline (e.g., 40.1 - 25.7 = 14.4 for Clipart), not gains over NP or Div. Please clarify the comparison baseline.
  2. [Sec. 4.2] The detector name 'RT-DTER' is a typo and should be 'RT-DETR'.
  3. [Sec. 3.3 / Algorithm 1] The symbol M is used both for image dimensions in Eq. (2) and for the number of GMM components (M=5) in Algorithm 1 and Sec. 3.3. Please use distinct symbols to avoid confusion.
  4. [Sec. 9.2] The limitation section refers to 'the more advanced detector YOLOv11' but cites reference [31], which is the Ultralytics YOLOv8 technical report; please provide the correct reference or paper for YOLOv11.
  5. [Supplementary Sec. 6, Table 6] Table 6 formatting is confusing: the first row '0.0 0.5' appears under the β columns, making it hard to identify which rows correspond to γ1, γ2 variations versus β1, β2 variations. Please reformat the table with clear row grouping or subheadings.

Circularity Check

1 steps flagged · score 4.0 of 10

Night-rainy gains are fitted, not predicted: the mixing hyperparameters are tuned on that test set and then reused for every reported result.

  1. fitted input called prediction [Sec. 6 (Parameter Sensitivity Analysis), Table 6; Sec. 7.2 (last paragraph); Sec. 4.2 (hyperparameters)]
    "As shown in Tab. 6, we analyze the sensitivity of the style mixing parameters in the proposed LDDS and provide quantitative results on the daytime-sunny to night-rainy task. ... a well-chosen higher mixing ratio is preferable (γ1 = 0.5, γ2 = 1.0). ... applying a proper bias towards the source mixed image feature yielded the best experimental results (β1 = 0.1, β2 = 2.0) ... our method uses identical parameters across all test datasets, without specific tuning for individual cases."

    The paper selects the mixing hyperparameters by maximizing mAP on the night-rainy test subset itself (Table 6), then fixes those exact values (γ1=0.5, γ2=1.0; β1=0.1, β2=2.0) in Sec. 4.2 and uses them to report the night-rainy results in Tables 2, 3, and 7. Thus the headline night-rainy improvement is not an independent prediction of the method: the parameters were tuned on the very metric and test domain being reported. The Sec. 7.2 claim that the method uses identical parameters without per-case tuning is literally true, but the single shared configuration was chosen on one of the test cases, so that domain's reported gain is statistically forced rather than a genuine generalization result.

full rationale

The core LDDS derivation is not circular: dual style mixing is a new augmentation pipeline built on CLIPstyler-style style generation, Fourier amplitude mixing, and feature-statistic mixing, and none of these steps are defined in terms of the evaluation results. I found no self-citation chain, no uniqueness theorem imported from the authors' prior work, and no ansatz smuggled in by self-citation. The only concrete circularity is empirical: the mixing hyperparameters are tuned on the night-rainy test set (Sec. 6, Table 6) and then used for all experiments, which makes the night-rainy gains a selected result rather than a prediction. Because the same fixed hyperparameters are also used for the other adverse-weather and cartoon benchmarks, those results are not as directly compromised, but they are not entirely independent either. A held-out validation split or a sensitivity analysis performed without test-set feedback would be needed to make the state-of-the-art claim clean. The style-prompt requirement (target domains must be nameable in advance) is a deployability limitation and a correctness risk, but not a circularity, so it does not affect this score.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The method introduces no new physical or conceptual entities. It relies heavily on prior work for each building block, but the central claim depends on several domain assumptions about CLIP and style transfer, plus five hyperparameters fitted to the night-rainy test set. The GMM smoothing is an ad hoc heuristic that improves empirical results.

free parameters (5)
  • gamma1 = 0.5
    Lower bound of the uniform range for image-level amplitude mixing weight; selected via sensitivity analysis on the night-rainy test set (Sec. 6).
  • gamma2 = 1.0
    Upper bound of the uniform range for image-level amplitude mixing; chosen to maximize night-rainy mAP in Sec. 6.
  • beta1 = 0.1
    First shape parameter of the Beta distribution for feature-level style mixing; selected on the night-rainy test set (Sec. 6).
  • beta2 = 2.0
    Second shape parameter of the Beta distribution; chosen by night-rainy test performance in Sec. 6.
  • M = 5
    Number of Gaussian components in the GMM smoothing step; fixed heuristic used in Algorithm 1, with no independent justification beyond the ablation.
assumptions (5)
  • domain assumption CLIP text embeddings are semantically aligned with visual style descriptions so that prompts can transfer style information to StyleNet.
    The entire style generation stage relies on the CLIP loss between text and image to guide StyleNet fine-tuning (Sec. 3.1).
  • standard math Fourier amplitude encodes global image style and phase encodes content.
    Used in Eqs. 2-4 for image-level style mixing, following prior work on Fourier domain augmentation [54, 56].
  • domain assumption Channel-wise mean and standard deviation of feature maps encode style information.
    Basis of feature-level style mixing, directly adopted from ADIN and MixStyle [28, 63].
  • ad hoc to paper A Gaussian Mixture Model with M=5 on feature statistics selects the most representative style information and avoids style conflicts.
    Introduced specifically in Sec. 3.3 and Fig. 4 to justify the smoothing step; supported only by the ablation, not by a deeper analysis.
  • domain assumption The target domain styles are known in advance and expressible by simple prompts.
    The prompt construction in Sec. 3.1 and Sec. 7.1 describes exactly the tested conditions (rain, night, foggy, cartoon), so the method requires this prior knowledge.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Language-Driven Dual Style Mixing for Single-Domain Generalized Object Detection." pith.science (2026). https://pith.science/paper/QHBM65L3

@misc{pith2026250507219,
  author       = {Pith},
  title        = {Pith review of: Language-Driven Dual Style Mixing for Single-Domain Generalized Object Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QHBM65L3}},
  note         = {Machine review of arXiv:2505.07219}
}
read the original abstract

Generalizing an object detector trained on a single domain to multiple unseen domains is a challenging task. Existing methods typically introduce image or feature augmentation to diversify the source domain to raise the robustness of the detector. Vision-Language Model (VLM)-based augmentation techniques have been proven to be effective, but they require that the detector's backbone has the same structure as the image encoder of VLM, limiting the detector framework selection. To address this problem, we propose Language-Driven Dual Style Mixing (LDDS) for single-domain generalization, which diversifies the source domain by fully utilizing the semantic information of the VLM. Specifically, we first construct prompts to transfer style semantics embedded in the VLM to an image translation network. This facilitates the generation of style diversified images with explicit semantic information. Then, we propose image-level style mixing between the diversified images and source domain images. This effectively mines the semantic information for image augmentation without relying on specific augmentation selections. Finally, we propose feature-level style mixing in a double-pipeline manner, allowing feature augmentation to be model-agnostic and can work seamlessly with the mainstream detector frameworks, including the one-stage, two-stage, and transformer-based detectors. Extensive experiments demonstrate the effectiveness of our approach across various benchmark datasets, including real to cartoon and normal to adverse weather tasks. The source code and pre-trained models will be publicly available at https://github.com/qinhongda8/LDDS.

Figures

Figures reproduced from arXiv: 2505.07219 by the authors.

Figure 1
Figure 1. Existing VLM-based methods [16, 49] (a) require that the detector’s backbone has the same architecture as the image en￾coder, resulting in the limitation of detector framework selection. Besides, they only perform feature augmentation of the detector’s backbone, can not effectively transfer the semantic information embedded in the VLM. Our method (b) avoids these limitations with proposed dual-style mixing schemes f… view at source ↗
Figure 2
Figure 2. The pipeline of LDDS, which consists of style generation, image-level and smooth feature-level style mixing. We first feed the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The visual results of the style generation and image-level [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: The semantic information space representation of [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Qualitative results on the real to cartoon. From left to right, every two columns represent the samples of Clipart1k, Watercolor2k, [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Qualitative results on the night-rainy dataset. We com [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Results for model performance on the night-rainy [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: The reported effect of style generation on the perfor [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: Illustration of the image augmentation visualization results, including Diversified [ [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: Visualization results on the target domain of daytime [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 11
Figure 11. Figure 11: Additional qualitative results on the target domain of daytime-sunny to night-rainy for YOLOv8 (source) [ [PITH_FULL_IMAGE:figures/full_fig_p015_11.png]
Figure 12
Figure 12. Figure 12: Additional qualitative results on the target domain of daytime-sunny to daytime-foggy for Faster R-CNN (source) [ [PITH_FULL_IMAGE:figures/full_fig_p016_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 58 canonical work pages

  1. [61]

    DETRs beat YOLOs on real-time object detection

    Yian Zhao, Wenyu Lv, Shangliang Xu, Jinman Wei, Guanzhong Wang, Qingqing Dang, Yi Liu, and Jie Chen. DETRs beat YOLOs on real-time object detection. InCVPR,

    work page

  2. [1]

    YOLOv4: Optimal speed and accuracy of object detection

    Alexey Bochkovskiy, Chien-Yao Wang, and Hong- Yuan Mark Liao. YOLOv4: Optimal speed and accuracy of object detection. arXiv preprint arXiv:2004.10934, 2020. 2

    arXiv 2004

  3. [2]

    End-to- end object detection with transformers

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. End-to- end object detection with transformers. In ECCV, 2020. 2

    2020

  4. [3]

    Harmonizing transferability and discriminability for adapting object detectors

    Chaoqi Chen, Zebiao Zheng, Xinghao Ding, Yue Huang, and Qi Dou. Harmonizing transferability and discriminability for adapting object detectors. In CVPR, 2020. 1

    work page 2020

  5. [4]

    Jieneng Chen, Shuyang Sun, Ju He, Philip H. S. Torr, Alan L. Yuille, and Song Bai. TransMix: Attend to mix for vision transformers. In CVPR, 2022. 3

    work page 2022

  6. [5]

    Meta-causal learning for single domain generalization

    Jin Chen, Zhi Gao, Xinxiao Wu, and Jiebo Luo. Meta-causal learning for single domain generalization. In CVPR, 2023. 1

    work page 2023

  7. [6]

    Learning domain adaptive object detection with probabilistic teacher

    Meilin Chen, Weijie Chen, Shicai Yang, Jie Song, Xin- chao Wang, Lei Zhang, Yunfeng Yan, Donglian Qi, Yueting Zhuang, Di Xie, and Shiliang Pu. Learning domain adaptive object detection with probabilistic teacher. In ICML, 2022. 1

    work page 2022

  8. [7]

    Center-aware adversarial augmentation for single domain generalization

    Tianle Chen, Mahsa Baktashmotlagh, Zijian Wang, and Mathieu Salzmann. Center-aware adversarial augmentation for single domain generalization. In WACV, 2023. 1

    work page 2023

Show all 68 references
  1. [8]

    Domain adaptive faster R-CNN for object detection in the wild

    Yuhua Chen, Wen Li, Christos Sakaridis, Dengxin Dai, and Luc Van Gool. Domain adaptive faster R-CNN for object detection in the wild. In CVPR, 2018. 1

    2018

  2. [9]

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

    Sungha Choi, Sanghun Jung, Huiwon Yun, Joanne T Kim, Seungryong Kim, and Jaegul Choo. RobustNet: Improving domain generalization in urban-scene segmentation via in- stance selective whitening. In CVPR, 2021. 6

    2021

  3. [10]

    Attention consistency on visual corruptions for single-source domain generalization

    Ilke Cugu, Massimiliano Mancini, Yanbei Chen, and Zeynep Akata. Attention consistency on visual corruptions for single-source domain generalization. In CVPR, 2022. 1, 8, 2, 3

    2022

  4. [11]

    Saquib Sarfraz, and Mohsen Ali

    Muhammad Sohail Danish, Muhammad Haris Khan, Muhammad Akhtar Munir, M. Saquib Sarfraz, and Mohsen Ali. Improving single domain-generalized object detection: A focus on diversification and alignment. In CVPR, 2024. 2, 6, 7

    2024

  5. [12]

    Un- biased mean teacher for cross-domain object detection

    Jinhong Deng, Wen Li, Yuhua Chen, and Lixin Duan. Un- biased mean teacher for cross-domain object detection. In CVPR, 2021. 1

    2021

  6. [13]

    D3T: Distinctive dual-domain teacher zigzagging across RGB- thermal gap for domain-adaptive object detection

    Dinh Phat Do, Taehoon Kim, Jaemin Na, Jiwon Kim, Keonho Lee, Kyunghwan Cho, and Wonjun Hwang. D3T: Distinctive dual-domain teacher zigzagging across RGB- thermal gap for domain-adaptive object detection. In CVPR,

  7. [14]

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

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition ...

    2021

  8. [15]

    The pascal visual object classes (VOC) challenge

    Mark Everingham, Luc Van Gool, Christopher KI Williams, John Winn, and Andrew Zisserman. The pascal visual object classes (VOC) challenge. IJCV, 2010. 5

    2010

  9. [16]

    PØDA: Prompt-driven zero- shot domain adaptation

    Mohammad Fahes, Tuan-Hung Vu, Andrei Bursuc, Patrick P´erez, and Raoul de Charette. PØDA: Prompt-driven zero- shot domain adaptation. In ICCV, 2023. 1, 2, 4, 6

    2023

  10. [17]

    Towards robust ob- ject detection invariant to real-world domain shifts

    Qi Fan, Mattia Seg `u, Yu-Wing Tai, Fisher Yu, Chi-Keung Tang, Bernt Schiele, and Dengxin Dai. Towards robust ob- ject detection invariant to real-world domain shifts. InICLR,

  11. [18]

    Adversarially adaptive normalization for single domain generalization

    Xinjie Fan, Qifei Wang, Junjie Ke, Feng Yang, Boqing Gong, and Mingyuan Zhou. Adversarially adaptive normalization for single domain generalization. In CVPR, 2021. 1

    2021

  12. [19]

    AcroFOD: An adaptive method for cross-domain few-shot object detection

    Yipeng Gao, Lingxiao Yang, Yunmu Huang, Song Xie, Shiy- ong Li, and Wei-Shi Zheng. AcroFOD: An adaptive method for cross-domain few-shot object detection. In ECCV, 2022. 1

    2022

  13. [20]

    AsyFOD: An asymmetric adaptation paradigm for few-shot domain adaptive object detection

    Yipeng Gao, Kun-Yu Lin, Junkai Yan, Yaowei Wang, and Wei-Shi Zheng. AsyFOD: An asymmetric adaptation paradigm for few-shot domain adaptive object detection. In CVPR, 2023. 1

    2023

  14. [21]

    Gatys, Alexander S

    Leon A. Gatys, Alexander S. Ecker, and Matthias Bethge. Image style transfer using convolutional neural networks. In CVPR, 2016. 4

    2016

  15. [22]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In CVPR,

  16. [23]

    Mask R-CNN

    Kaiming He, Georgia Gkioxari, Piotr Doll ´ar, and Ross Gir- shick. Mask R-CNN. In ICCV, 2017. 2, 3

    2017

  17. [24]

    Multi-adversarial faster-rcnn for unrestricted object detection

    Zhenwei He and Lei Zhang. Multi-adversarial faster-rcnn for unrestricted object detection. In ICCV, 2019. 1

    2019

  18. [25]

    StyleMix: Sep- arating content and style for enhanced data augmentation

    Minui Hong, Jinwoo Choi, and Gunhee Kim. StyleMix: Sep- arating content and style for enhanced data augmentation. In CVPR, 2021. 3

    2021

  19. [26]

    Mixed samples as probes for unsu- pervised model selection in domain adaptation

    Dapeng Hu, Jian Liang, Jun Hao Liew, Chuhui Xue, Song Bai, and Xinchao Wang. Mixed samples as probes for unsu- pervised model selection in domain adaptation. In NeurIPS,

  20. [27]

    Itera- tive normalization: Beyond standardization towards efficient whitening

    Lei Huang, Yi Zhou, Fan Zhu, Li Liu, and Ling Shao. Itera- tive normalization: Beyond standardization towards efficient whitening. In CVPR, 2019. 6

    2019

  21. [28]

    Arbitrary style transfer in real-time with adaptive instance normalization

    Xun Huang and Serge Belongie. Arbitrary style transfer in real-time with adaptive instance normalization. In ICCV,

  22. [29]

    Cross-domain weakly-supervised object de- tection through progressive domain adaptation

    Naoto Inoue, Ryosuke Furuta, Toshihiko Yamasaki, and Kiy- oharu Aizawa. Cross-domain weakly-supervised object de- tection through progressive domain adaptation. In CVPR,

  23. [30]

    Diffusemix: Label- preserving data augmentation with diffusion models

    Khawar Islam, Muhammad Zaigham Zaheer, Arif Mah- mood, and Karthik Nandakumar. Diffusemix: Label- preserving data augmentation with diffusion models. In CVPR, 2024. 3

    2024

  24. [31]

    Ultralytics YOLOv8, 2023

    Glenn Jocher, Ayush Chaurasia, and Jing Qiu. Ultralytics YOLOv8, 2023. 2, 6, 7, 8, 3, 4, 5

    2023

  25. [32]

    Driving in the matrix: Can virtual worlds replace human- generated annotations for real world tasks? arXiv preprint arXiv:1610.01983, 2016

    Matthew Johnson-Roberson, Charles Barto, Rounak Mehta, Sharath Nittur Sridhar, Karl Rosaen, and Ram Vasudevan. Driving in the matrix: Can virtual worlds replace human- generated annotations for real world tasks? arXiv preprint arXiv:1610.01983, 2016. 1

    arXiv

  26. [33]

    Mikhail Kennerley, Jian-Gang Wang, Bharadwaj Veeravalli, and Robby T. Tan. CAT: Exploiting inter-class dynamics for domain adaptive object detection. In CVPR, 2024. 1

    2024

  27. [34]

    Puz- zle mix: Exploiting saliency and local statistics for optimal mixup

    Jang-Hyun Kim, Wonho Choo, and Hyun Oh Song. Puz- zle mix: Exploiting saliency and local statistics for optimal mixup. In ICML, 2020. 3

    2020

  28. [35]

    CLIPstyler: Image style transfer with a single text condition

    Gihyun Kwon and Jong Chul Ye. CLIPstyler: Image style transfer with a single text condition. In CVPR, 2022. 2, 3, 4

    2022

  29. [36]

    Reed, Cheng-Yang Fu, and Alexander C

    Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott E. Reed, Cheng-Yang Fu, and Alexander C. Berg. SSD: Single shot MultiBox detector. In ECCV, 2016. 2

    2016

  30. [37]

    ConfMix: Unsupervised domain adaptation for ob- ject detection via confidence-based mixing

    Giulio Mattolin, Luca Zanella, Elisa Ricci, and Yiming Wang. ConfMix: Unsupervised domain adaptation for ob- ject detection via confidence-based mixing. In CVPR, 2023. 3

    2023

  31. [38]

    Nussbaumer and Henri J

    Henri J. Nussbaumer and Henri J. Nussbaumer. The fast Fourier transform. Springer, 1982. 4

    1982

  32. [39]

    DoubleAUG: Single-domain generalized object detector in urban via color perturbation and dual-style memory.TOMM,

    Lei Qi, Peng Dong, Tan Xiong, Hui Xue, and Xin Geng. DoubleAUG: Single-domain generalized object detector in urban via color perturbation and dual-style memory.TOMM,

  33. [40]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision. In ICML, 2021. 2

    2021

  34. [41]

    YOLO9000: Better, faster, stronger

    Joseph Redmon and Ali Farhadi. YOLO9000: Better, faster, stronger. In CVPR, 2017. 2, 3

    2017

  35. [42]

    You only look once: Unified, real-time object de- tection

    Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. You only look once: Unified, real-time object de- tection. In CVPR, 2016. 1, 2, 3

    2016

  36. [43]

    Faster R-CNN: Towards real-time object detection with re- gion proposal networks

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster R-CNN: Towards real-time object detection with re- gion proposal networks. In NeurIPS, 2015. 1, 2, 6, 3

    2015

  37. [44]

    U- Net: Convolutional networks for biomedical image segmen- tation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U- Net: Convolutional networks for biomedical image segmen- tation. In MICCAI, 2015. 4

    2015

  38. [45]

    Se- mantic foggy scene understanding with synthetic data.IJCV,

    Christos Sakaridis, Dengxin Dai, and Luc Van Gool. Se- mantic foggy scene understanding with synthetic data.IJCV,

  39. [46]

    FCOS: A simple and strong anchor-free object detector

    Zhi Tian, Chunhua Shen, Hao Chen, and Tong He. FCOS: A simple and strong anchor-free object detector. TPAMI, 2022. 2

    2022

  40. [47]

    DACS: Domain adaptation via cross- domain mixed sampling

    Wilhelm Tranheden, Viktor Olsson, Juliano Pinto, and Lennart Svensson. DACS: Domain adaptation via cross- domain mixed sampling. In WCAV, 2021. 3

    2021

  41. [48]

    Manifold mixup: Better representations by interpolating hidden states

    Vikas Verma, Alex Lamb, Christopher Beckham, Amir Na- jafi, Ioannis Mitliagkas, David Lopez-Paz, and Yoshua Ben- gio. Manifold mixup: Better representations by interpolating hidden states. In ICML, 2019. 3

    2019

  42. [49]

    CLIP the gap: A single domain generalization approach for object detection

    Vidit Vidit, Martin Engilberge, and Mathieu Salzmann. CLIP the gap: A single domain generalization approach for object detection. In CVPR, 2023. 1, 2, 4, 6

    2023

  43. [50]

    Few- shot adaptive faster R-CNN

    Tao Wang, Xiaopeng Zhang, Li Yuan, and Jiashi Feng. Few- shot adaptive faster R-CNN. In CVPR, 2019. 1

    2019

  44. [51]

    Learning to diversify for single do- main generalization

    Zijian Wang, Yadan Luo, Ruihong Qiu, Zi Huang, and Mahsa Baktashmotlagh. Learning to diversify for single do- main generalization. In ICCV, 2021. 1

    2021

  45. [52]

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

    Aming Wu and Cheng Deng. Single-domain generalized object detection in urban scene via cyclic-disentangled self- distillation. In CVPR, 2022. 2, 6, 3

    2022

  46. [53]

    Vector-decomposed disentanglement for domain- invariant object detection

    Aming Wu, Rui Liu, Yahong Han, Linchao Zhu, and Yi Yang. Vector-decomposed disentanglement for domain- invariant object detection. In ICCV, 2021. 1

    2021

  47. [54]

    A fourier-based framework for domain generaliza- tion

    Qinwei Xu, Ruipeng Zhang, Ya Zhang, Yanfeng Wang, and Qi Tian. A fourier-based framework for domain generaliza- tion. In CVPR, 2021. 4, 8, 2, 3

    2021

  48. [55]

    SimDE: A simple domain expan- sion approach for single-source domain generalization

    Qinwei Xu, Ruipeng Zhang, Yi-Yan Wu, Ya Zhang, Ning Liu, and Yanfeng Wang. SimDE: A simple domain expan- sion approach for single-source domain generalization. In CVPRW, 2023. 1

    2023

  49. [56]

    FDA: Fourier domain adaptation for semantic segmentation

    Yanchao Yang and Stefano Soatto. FDA: Fourier domain adaptation for semantic segmentation. In CVPR, 2020. 4, 8, 2, 3

    2020

  50. [57]

    Zou, and Chelsea Finn

    Huaxiu Yao, Yiping Wang, Linjun Zhang, James Y . Zou, and Chelsea Finn. C-Mixup: Improving generalization in regres- sion. In NeurIPS, 2022. 3

    2022

  51. [58]

    BDD100K: A diverse driving dataset for heterogeneous multitask learning

    Fisher Yu, Haofeng Chen, Xin Wang, Wenqi Xian, Yingying Chen, Fangchen Liu, Vashisht Madhavan, and Trevor Dar- rell. BDD100K: A diverse driving dataset for heterogeneous multitask learning. In CVPR, 2020. 6

    2020

  52. [59]

    CutMix: Regu- larization strategy to train strong classifiers with localizable features

    Sangdoo Yun, Dongyoon Han, Seong Joon Oh, Sanghyuk Chun, Junsuk Choe, and Youngjoon Yoo. CutMix: Regu- larization strategy to train strong classifiers with localizable features. In CVPR, 2019. 3

    2019

  53. [60]

    Few-shot cross-domain object detection with instance- level prototype-based meta-learning

    Lin Zhang, Bo Zhang, Botian Shi, Jiayuan Fan, and Tao Chen. Few-shot cross-domain object detection with instance- level prototype-based meta-learning. TCSVT, 2024. 1

    2024

  54. [62]

    PICA: Point-wise instance and cen- troid alignment based few-shot domain adaptive object de- tection with loose annotations

    Chaoliang Zhong, Jie Wang, Cheng Feng, Ying Zhang, Jun Sun, and Yasuto Yokota. PICA: Point-wise instance and cen- troid alignment based few-shot domain adaptive object de- tection with loose annotations. In WACV, 2022. 1

    2022

  55. [63]

    Mixstyle neural networks for domain generalization and adaptation

    Kaiyang Zhou, Yongxin Yang, Yu Qiao, and Tao Xiang. Mixstyle neural networks for domain generalization and adaptation. IJCV, 2024. 3, 4, 5, 8

    2024

  56. [64]

    Patch-mix trans- former for unsupervised domain adaptation: A game per- spective

    Jinjing Zhu, Haotian Bai, and Lin Wang. Patch-mix trans- former for unsupervised domain adaptation: A game per- spective. In CVPR, 2023. 3 In the supplementary material, Sec. 6 begins with a sen- sitivity analysis of the two mixing parameters in dual-style mixing. Sec. 7 provi...

    2023

  57. [65]

    6, we analyze the sensitivity of the style mixing parameters in the proposed LDDS and pro- vide quantitative results on the daytime-sunny to night- rainy task

    Parameter Sensitivity Analysis As shown in Tab. 6, we analyze the sensitivity of the style mixing parameters in the proposed LDDS and pro- vide quantitative results on the daytime-sunny to night- rainy task. For the image-level style mixing parameter ω ∼ U(γ1,γ 2) and feature-...

  58. [66]

    Implementation Details In the experiments based on one-stage, two-stage, and transformer-based detectors, the style generation compo- nent adhered to a consistent setup

    Additional Experimental Details 7.1. Implementation Details In the experiments based on one-stage, two-stage, and transformer-based detectors, the style generation compo- nent adhered to a consistent setup. Our style generation module, composed of a ResNet-101 CLIP image encod...

  59. [67]

    Additional Visualization Results 8.1. Image augmentation visualization results We analyze the effects of image augmentation on SDG object detection by providing visual comparisons of dif- ferent augmentation methods on source images, including Fourier Corruption [10], Fourier ...

  60. [68]

    Structure of the backbone The backbone network serving as the feature extraction module is a critical component of the detector

    Discussion 9.1. Structure of the backbone The backbone network serving as the feature extraction module is a critical component of the detector. Its struc- ture depends not only on the parameter size but also on the architecture type of the detector. For two-stage detec- tors ...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.