Pith. sign in

REVIEW 4 major objections 6 minor 51 references

PQ-DAF: Pose-driven Quality-controlled Data Augmentation for Data-scarce Driver Distraction Detection

T0 review · 4 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Pose-conditioned diffusion plus a vision-language quality filter is claimed to lift few-shot driver distraction detection from 36.67% to 54.00% at 10 shots and from 64.67% to 88.00% at 30 shots on StateFarm.

desk verdict Plausible idea, but the few-shot protocol leaks the training distribution into the classifier and the numbers are inconsistent, so the headline gains are not established. read the letter →

arxiv 2508.10397 v1 pith:2CBNKJ6N submitted 2025-08-14 cs.CV cs.AI

Haibin Sun , Xinghui Song This is my paper
classification cs.CVcs.AI
keywords driverdistractiondetectionfew-shotlearningpose-guidedimagesynthesisdiffusionmodelsdataaugmentationvision-languagemodelqualityfilteringdomainshift
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

PQ-DAF claims that a two-stage pipeline—pose-conditioned diffusion generation followed by vision-language-model filtering—can turn a handful of labeled driver images into a much larger, reliable training set. The paper argues that because distraction behaviors are strongly tied to body pose, synthesizing new images under extracted keypoint poses preserves the action semantics that matter, while a CogVLM confidence threshold removes samples whose action is ambiguous. On the StateFarm and AUC-DDD benchmarks, coupling this augmentation with ResNet50 raises 10-shot accuracy from 36.67% to 54.00% and 30-shot from 64.67% to 88.00%. If correct, the method offers a way to build robust in-vehicle distraction detectors without large-scale manual annotation.

What carries the argument

The central machinery is PCDMs (Progressive Conditional Diffusion Models), a pose-guided latent diffusion generator whose input fuses three condition branches—source image plus mask, paired source-target images, and source-target pose maps—encoded through a frozen DINOv2 encoder and a pose encoder, with classifier-free guidance combining image and pose branches during inference. The secondary machinery is the CogVLM-based quality gate: a numerical image-text matching score in $[0,1]$ against a per-category prompt, with a fixed threshold $\tau = 0.8$ deciding which synthetic samples enter the training set. Together they enforce both structural pose consistency and semantic action purity.

What would settle it

Train the PCDM generator on exactly the same 10-shot or 30-shot subset used for the classifier, instead of the 80% training split. If the accuracy gain over real-only training collapses or reverses, the reported improvement is explained by the generator having seen the evaluation distribution, not by pose-guided augmentation. Alternatively, hold out entire subjects from the generator's training set and evaluate only on those subjects; if gains vanish, the method has not generalized.

Watch

Extended reading notes

Core claim

The central claim is that a progressive conditional diffusion model, conditioned on DWpose keypoint maps and source-image appearance, can generate driver-distraction images that are structurally consistent with the target pose and semantically consistent with the intended action; the CogVLM-based scoring function $s = C(I_g, \text{Prompt}[c])$ with threshold $\tau = 0.8$ then discards generated samples whose image-text match is weak. The surviving pseudo-samples are mixed with real few-shot data and used to train a classifier. The paper reports that this pipeline improves Top-1 accuracy by 17.33 percentage points at 10-shot and 23.33 points at 30-shot on StateFarm, and by 10 and 20.66 points

Load-bearing premise

The gains count only if using test-set images as few-shot training data, alongside synthetic images from a generator trained on the training set, still measures generalization to new drivers rather than leakage of the training distribution.

Editorial extensions

If this is right

  • PQ-DAF offers a path to expand scarce driver-distraction datasets without manual annotation, because the quality filter is automatic and per-category prompts are the only human input.
  • Real-to-synthetic ratio is a practical lever: medium and large models improve as the ratio rises to 1:3, while lightweight models peak near 1:1–1:2 and degrade with too much synthetic data.
  • The generation-model ablation indicates that identity- and pose-consistent synthesis (PCDMs) matters more than pose-only conditioning (ControlNet, IP-Adapter), giving a concrete design rule for future augmentation frameworks.
  • The semantic filter removes action-ambiguous samples, which the paper argues is why PQ-DAF beats pixel-mixing augmentations such as Mixup, CutMix, and PuzzleMix in few-shot settings.

Reading between the lines

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

  • Beyond the paper's claims: the same pose-conditioned generation plus automatic semantic filtering could transfer to other few-shot fine-grained action-recognition tasks where pose strongly determines the label, such as in-cabin monitoring or assembly-line safety.
  • The fixed threshold $\tau = 0.8$ is never swept; an adaptive or category-specific threshold would be a natural testable extension to see whether the quality-retention trade-off can be improved.
  • The paper leaves implicit that synthetic augmentation could act as a privacy-preserving way to share driving behavior data, since raw in-vehicle footage need not be released, although synthetic images may still encode identity cues.
  • Because gains concentrate in larger-capacity models, an untested extension is to pair PQ-DAF with distillation into lightweight models, which are the realistic deployment target in vehicles.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes PQ-DAF, a data-augmentation framework for few-shot driver distraction detection. It combines pose extraction with DWpose, pose-conditioned synthesis with Progressive Conditional Diffusion Models (PCDMs), and semantic filtering with the CogVLM vision-language model at a fixed confidence threshold. The augmented synthetic samples are mixed with real few-shot samples at various ratios and used to train standard classifiers. Experiments on StateFarm and AUC-DDD in 10-shot and 30-shot settings report large Top-1 accuracy gains over baselines, e.g., ResNet50 from 36.67% to 54.00% on StateFarm 10-shot and from 50.67% to 71.33% on AUC-DDD 30-shot.

Significance. If the results are valid, the idea is practically useful: pose-conditioned diffusion with VLM-based quality filtering is a sensible approach to data-scarce driver distraction detection, and the paper builds on established components (PCDMs, CogVLM, DINOv2) without introducing circular derivations in the equations. Equations (1) and (2) are standard conditional diffusion objectives, and the pipeline uses external pretrained models, which is a strength. However, the experimental section as written does not establish the claimed generalization gains: the few-shot training subsets are drawn from the original test split while the generator is trained on the original training split, and the evaluation split is never specified. The reported improvements could therefore reflect split leakage rather than the proposed augmentation. Unreconciled numerical inconsistencies across tables further weaken the empirical support.

major comments (4)
  1. [§IV-A, Tables II–III] The protocol says the 10/30-shot subsets are randomly sampled from the original test split, while the PCDMs generator is fine-tuned on the original training split, and the paper never states where the reported Top-1 accuracy is measured. If evaluation uses the remaining original test images, the few-shot classifier training images come from the same distribution (and likely same subjects) as the test set; if evaluation uses the original training split, the generator has been trained on the evaluation distribution. Either reading contaminates the comparison, and the claimed gains (e.g., 36.67→54.00 in Table II; 30.67→40.67 and 50.67→71.33 in Table III) are not attributable to pose conditioning or filtering. Please define a held-out evaluation split disjoint from both generator training and few-shot classifier training and report all numbers under that protocol.
  2. [§IV-D1/Table II vs §IV-D3/Table IV; Conclusion] The same 10-shot StateFarm result for OURS is 54.00 in Table II but 54.67 in Table IV. The conclusion reports an Inceptionv4 AUC-DDD gain 'from 19.33% to 34.00%,' but 34.00% does not appear in Table III, whose OURS 10-shot value is 40.67%. These discrepancies make the reported empirical base unreliable; a single consistent set of results must be presented and reconciled.
  3. [Abstract, §IV-A] The paper claims improvements in 'model generalization' and 'cross-domain robustness,' but no cross-dataset experiment is performed. On both StateFarm and AUC-DDD, the few-shot setting is a random within-dataset subsample; there is no train-on-one-domain/evaluate-on-another protocol. The abstract's domain-shift claim is therefore unsupported. A cross-dataset evaluation, e.g., training on one dataset's clean split and testing on the other, is needed to support the claimed generalization.
  4. [§III-C, Table VI] The contribution of the CogVLM filtering mechanism is never isolated. Table VI compares generation backbones, but all variants include the same filtering; no ablation removes Algorithm 1 or varies the threshold τ (fixed at 0.8). Since the title and contributions emphasize quality control, the experiments need to show that filtering itself, not merely added synthetic volume, is responsible for any observed gain.
minor comments (6)
  1. [Throughout] Terminology is inconsistent: PCDMs vs PCDMS, PQ-DAF vs PQ-ADF vs PQE-AD appear in §IV-D4, §IV-E, and Figures 4–6. Please standardize.
  2. [Table IV] Column header 'Statefarm' should be 'StateFarm'; use 'AUC-DDD' consistently instead of 'AUC-DDD'/'AucDDD'.
  3. [§IV-C] The PCDM fine-tuning is described as '30,000 epochs'; this is likely iterations or steps. Please clarify the exact training protocol.
  4. [References] References [8] and [28] are the same paper; merge or differentiate. Also [8] in the introduction cites an arXiv version while [28] cites the ICLR version.
  5. [Algorithm 1] The query string contains stray backslashes and the line is hard to parse; the pseudo-code format should be cleaned.
  6. [Tables II–III] No variance or random-seed information is reported for the main results; add mean±std over multiple seeds to support the 10/30-shot claims.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the augmentation pipeline is self-contained and no predicted quantity reduces to a fitted input or self-citation.

full rationale

The paper's derivation chain is empirical rather than deductive. The pose-conditioned generation loss (Eq. 1) and classifier-free guidance (Eq. 2) define a standard diffusion training objective; neither defines the reported accuracy in terms of the input. PCDMs and CogVLM are external models cited as prior work, not results derived in this paper, and the authors do not import a uniqueness theorem or adopt an ansatz via self-citation. Top-1 accuracy (Sec. IV-B) is a conventional metric computed on ground-truth labels. The few-shot protocol in Sec. IV-A draws 10/30-shot subsets from the original test split and fine-tunes PCDMs on the original training split; while the paper does not state the evaluation split, this is a potential data-leakage / experimental-validity concern, not a circular equivalence: no equation or fitted parameter forces the reported gain. Apparent self-citation [10] (H. Sun) is background statistics and not load-bearing. Therefore no circularity score is warranted.

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

The central claim rests on task-level assumptions (pose as label carrier, VLM score as label-correctness proxy, prompt-category alignment) and on tunable hyperparameters (threshold, guidance weight, mixture ratio) whose values are either fixed without analysis or selected post hoc. The paper does not introduce new theoretical entities.

free parameters (4)
  • CogVLM filtering threshold tau = 0.8
    Fixed threshold in Sec III-C-3; no sensitivity analysis is reported, and the retention-rate versus purity trade-off is asserted without data.
  • Classifier-free guidance weight w = not reported
    Appears in Eq. (2) and controls pose versus semantic branch fusion; the text only gives w in [0,1] and never states the value used in experiments.
  • Real-to-synthetic mixture ratio = 1:0.5 to 1:3, best selected per model
    Section IV-D-4 tunes the ratio after observing results; Table V shows large sensitivity, and the headline results in Tables II and III never state which ratio was used.
  • PCDM fine-tuning epochs = 30,000
    Reported in Sec IV-C; likely a typo or unreasonably large, and it is a training choice not derived from any principle.
assumptions (4)
  • domain assumption Pose is a sufficient structural prior for driver distraction actions.
    The whole generation pipeline conditions on DWpose keypoints and assumes synthesized images preserve the semantics of the source action (Sec III-B). If pose does not carry distinguishing features such as a phone in the hand, labels can be wrong.
  • domain assumption CogVLM's prompted confidence score is a valid measure of image-text semantic match and correlates with classification label correctness.
    Sec III-C uses the VLM score as the quality filter; no calibration or validation against human labels is provided.
  • domain assumption The 10 StateFarm and AUC-DDD categories map one-to-one to the 10 hand-written prompts in Table I.
    Each class gets a single prompt without checking ambiguities, such as the visual similarity between 'holding phone to right ear' and 'texting with right hand'.
  • standard math The standard denoising objective in Eq. (1) is a valid training target.
    This is the standard diffusion loss, inherited from PCDMs and the diffusion literature.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PQ-DAF: Pose-driven Quality-controlled Data Augmentation for Data-scarce Driver Distraction Detection." pith.science (2026). https://pith.science/paper/2CBNKJ6N

@misc{pith2026250810397,
  author       = {Pith},
  title        = {Pith review of: PQ-DAF: Pose-driven Quality-controlled Data Augmentation for Data-scarce Driver Distraction Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2CBNKJ6N}},
  note         = {Machine review of arXiv:2508.10397}
}
read the original abstract

Driver distraction detection is essential for improving traffic safety and reducing road accidents. However, existing models often suffer from degraded generalization when deployed in real-world scenarios. This limitation primarily arises from the few-shot learning challenge caused by the high cost of data annotation in practical environments, as well as the substantial domain shift between training datasets and target deployment conditions. To address these issues, we propose a Pose-driven Quality-controlled Data Augmentation Framework (PQ-DAF) that leverages a vision-language model for sample filtering to cost-effectively expand training data and enhance cross-domain robustness. Specifically, we employ a Progressive Conditional Diffusion Model (PCDMs) to accurately capture key driver pose features and synthesize diverse training examples. A sample quality assessment module, built upon the CogVLM vision-language model, is then introduced to filter out low-quality synthetic samples based on a confidence threshold, ensuring the reliability of the augmented dataset. Extensive experiments demonstrate that PQ-DAF substantially improves performance in few-shot driver distraction detection, achieving significant gains in model generalization under data-scarce conditions.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

51 extracted references · 39 canonical work pages

  1. [1]

    Driver distraction detection based on lightweight networks and tiny object detection,

    Z. Zhu, S. Wang, S. Gu, Y . Li, J. Li, L. Shuai, and G. Qi, “Driver distraction detection based on lightweight networks and tiny object detection,” Mathematical bio- sciences and engineering , vol. 20, no. 10, pp. 18 248– 18 266, 2023

    work page 2023

  2. [2]

    An intelligent network framework for driver distraction monitoring based on res- se-cnn,

    J. Lei, Z. Ni, Z. Peng, H. Hu, J. Hong, X. Fang, C. Yi, C. Ren, and M. A. Wasaye, “An intelligent network framework for driver distraction monitoring based on res- se-cnn,” Scientific Reports, vol. 15, no. 1, p. 6916, 2025

    work page 2025

  3. [3]

    Training convolutional neural networks with class based data augmentation for detecting distracted drivers,

    J. Cronje and A. P. Engelbrecht, “Training convolutional neural networks with class based data augmentation for detecting distracted drivers,” in Proceedings of the 9th International Conference on Computer and Automation Engineering, 2017, pp. 126–130

    work page 2017

  4. [4]

    Analysis of traffic accident causes based on data augmentation and ensemble learning with high-dimensional small-sample data,

    L. Zhu, Z. Zhang, D. Song, and B. Chen, “Analysis of traffic accident causes based on data augmentation and ensemble learning with high-dimensional small-sample data,” Expert Systems with Applications , vol. 237, p. 121782, 2024

    work page 2024

  5. [5]

    A transfer learning approach for identification of distracted driving,

    I. K. K. Ugli, A. Hussain, B. S. Kim, S. Aich, and H.-C. Kim, “A transfer learning approach for identification of distracted driving,” in 2022 24th International Confer- ence on Advanced Communication Technology (ICACT) . IEEE, 2022, pp. 420–423

    work page 2022

  6. [6]

    A data augmen- tation approach to distracted driving detection,

    J. Wang, Z. Wu, F. Li, and J. Zhang, “A data augmen- tation approach to distracted driving detection,” Future internet, vol. 13, no. 1, p. 1, 2020

    work page 2020

  7. [7]

    Vision-language models can identify distracted driver behavior from naturalistic videos,

    M. Z. Hasan, J. Chen, J. Wang, M. S. Rahman, A. Joshi, S. Velipasalar, C. Hegde, A. Sharma, and S. Sarkar, “Vision-language models can identify distracted driver behavior from naturalistic videos,” IEEE Transactions on Intelligent Transportation Systems, vol. 25, no. 9, pp. 11 602–11 616, 2024

    work page 2024

  8. [8]

    Advancing pose-guided image synthesis with progressive conditional diffusion models,

    F. Shen, H. Ye, J. Zhang, C. Wang, X. Han, and W. Yang, “Advancing pose-guided image synthesis with progressive conditional diffusion models,” arXiv preprint arXiv:2310.06313, 2023

    arXiv 2023

Show all 51 references
  1. [9]

    Cogvlm: Visual expert for pretrained language models,

    W. Wang, Q. Lv, W. Yu, W. Hong, J. Qi, Y . Wang, J. Ji, Z. Yang, L. Zhao, S. XiXuan et al. , “Cogvlm: Visual expert for pretrained language models,” Advances in Neural Information Processing Systems , vol. 37, pp. 121 475–121 499, 2024

    2024

  2. [10]

    A review of research on driving distraction based on bibliometrics and co-occurrence: Focus on driving distraction recognition methods,

    H. Ge, Y . Bo, H. Sun, M. Zheng, and Y . Lu, “A review of research on driving distraction based on bibliometrics and co-occurrence: Focus on driving distraction recognition methods,” Journal of safety research , vol. 82, pp. 261– 10 274, 2022

    2022

  3. [11]

    Studying the effects of driver distraction and traffic density on the probability of crash and near-crash events in naturalistic driving environment,

    R. Tian, L. Li, M. Chen, Y . Chen, and G. J. Witt, “Studying the effects of driver distraction and traffic density on the probability of crash and near-crash events in naturalistic driving environment,” IEEE Transactions on Intelligent Transportation Systems, vol. 14, no. 3, p...

    2013

  4. [12]

    Driver distracted be- havior detection technology with yolo-based deep learn- ing networks,

    Y .-S. Poon, C.-Y . Kao, Y .-K. Wang, C.-C. Hsiao, M.-Y . Hung, Y .-C. Wang, and C.-P. Fan, “Driver distracted be- havior detection technology with yolo-based deep learn- ing networks,” in 2021 IEEE International Symposium on Product Compliance Engineering-Asia (ISPCE-ASIA). I...

    2021

  5. [13]

    Distracted driver detection using convolutional neural networks based segmentation model,

    A. Khellal, M. Boulahmar, A. Bahi, and A. Nemra, “Distracted driver detection using convolutional neural networks based segmentation model,” in 2024 2nd In- ternational Conference on Electrical Engineering and Automatic Control (ICEEAC). IEEE, 2024, pp. 1–6

    2024

  6. [14]

    Distracted driver detection based on a cnn with decreasing filter size,

    B. Qin, J. Qian, Y . Xin, B. Liu, and Y . Dong, “Distracted driver detection based on a cnn with decreasing filter size,” IEEE transactions on intelligent transportation systems, vol. 23, no. 7, pp. 6922–6933, 2021

    2021

  7. [15]

    Cat-capsnet: A convolutional and attention based capsule network to detect the driver’s distraction,

    H. Mittal and B. Verma, “Cat-capsnet: A convolutional and attention based capsule network to detect the driver’s distraction,” IEEE Transactions on Intelligent Trans- portation Systems, vol. 24, no. 9, pp. 9561–9570, 2023

    2023

  8. [16]

    Using naturalistic driving data to assess the prevalence of environmental factors and driver behaviors in teen driver crashes,

    C. Carney, D. McGehee, K. Harland, M. Weiss, and M. Raby, “Using naturalistic driving data to assess the prevalence of environmental factors and driver behaviors in teen driver crashes,” 2015

    2015

  9. [17]

    Driver distraction behavior detection using a vision transformer model based on transfer learning strategy,

    Z. Fang, J. Chen, J. Wang, Z. Wang, N. Liu, and G. Yin, “Driver distraction behavior detection using a vision transformer model based on transfer learning strategy,” in 2022 6th CAA International Conference on Vehicular Control and Intelligence (CVCI) . IEEE, 2022, pp. 1–6

    2022

  10. [18]

    A lightweight and efficient distracted driver detection model fusing convolutional neural network and vision trans- former,

    Z. Li, X. Zhao, F. Wu, D. Chen, and C. Wang, “A lightweight and efficient distracted driver detection model fusing convolutional neural network and vision trans- former,” IEEE Transactions on Intelligent Transportation Systems, 2024

    2024

  11. [19]

    Birswint: Bilinear full- scale residual swin-transformer for fine-grained driver behavior recognition,

    W. Yang, C. Tan, Y . Chen, H. Xia, X. Tang, Y . Cao, W. Zhou, L. Lin, and G. Dai, “Birswint: Bilinear full- scale residual swin-transformer for fine-grained driver behavior recognition,” Journal of the Franklin Institute , vol. 360, no. 2, pp. 1166–1183, 2023

    2023

  12. [20]

    Conditional generative ad- versarial nets,

    M. Mirza and S. Osindero, “Conditional generative ad- versarial nets,” arXiv preprint arXiv:1411.1784 , 2014

    2014 arXiv

  13. [21]

    Deformable gans for pose-based human image gen- eration,

    A. Siarohin, E. Sangineto, S. Lathuiliere, and N. Sebe, “Deformable gans for pose-based human image gen- eration,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 3408– 3416

    2018

  14. [22]

    Adain-based tunable cyclegan for efficient unsupervised low-dose ct denoising,

    J. Gu and J. C. Ye, “Adain-based tunable cyclegan for efficient unsupervised low-dose ct denoising,” IEEE Transactions on Computational Imaging , vol. 7, pp. 73– 85, 2021

    2021

  15. [23]

    Long-term talkingface generation via motion-prior conditional diffusion model,

    F. Shen, C. Wang, J. Gao, Q. Guo, J. Dang, J. Tang, and T.-S. Chua, “Long-term talkingface generation via motion-prior conditional diffusion model,” arXiv preprint arXiv:2502.09533, 2025

    2025 arXiv

  16. [24]

    Imaggarment-1: Fine-grained garment gener- ation for controllable fashion design,

    F. Shen, J. Yu, C. Wang, X. Jiang, X. Du, and J. Tang, “Imaggarment-1: Fine-grained garment gener- ation for controllable fashion design,” arXiv preprint arXiv:2504.13176, 2025

    2025 arXiv

  17. [25]

    Imagharmony: Controllable image editing with consistent object quantity and layout,

    F. Shen, X. Du, Y . Gao, J. Yu, Y . Cao, X. Lei, and J. Tang, “Imagharmony: Controllable image editing with consistent object quantity and layout,” arXiv preprint arXiv:2506.01949, 2025

    2025

  18. [26]

    Person image synthesis via denoising diffusion model,

    A. K. Bhunia, S. Khan, H. Cholakkal, R. M. Anwer, J. Laaksonen, M. Shah, and F. S. Khan, “Person image synthesis via denoising diffusion model,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 5968–5976

    2023

  19. [27]

    Multimodal garment designer: Human-centric latent diffusion models for fash- ion image editing,

    A. Baldrati, D. Morelli, G. Cartella, M. Cornia, M. Bertini, and R. Cucchiara, “Multimodal garment designer: Human-centric latent diffusion models for fash- ion image editing,” in Proceedings of the IEEE/CVF international conference on computer vision , 2023, pp. 23 393–23 402

    2023

  20. [28]

    Advancing pose-guided image synthesis with progressive conditional diffusion models,

    F. Shen, H. Ye, J. Zhang, C. Wang, X. Han, and Y . Wei, “Advancing pose-guided image synthesis with progressive conditional diffusion models,” in The Twelfth International Conference on Learning Representations ,

  21. [29]

    Humansd: A native skeleton-guided diffusion model for human image generation,

    X. Ju, A. Zeng, C. Zhao, J. Wang, L. Zhang, and Q. Xu, “Humansd: A native skeleton-guided diffusion model for human image generation,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 15 988–15 998

    2023

  22. [30]

    Imagpose: A unified conditional framework for pose-guided person generation,

    F. Shen and J. Tang, “Imagpose: A unified conditional framework for pose-guided person generation,” Advances in neural information processing systems , vol. 37, pp. 6246–6266, 2024

    2024

  23. [31]

    Imagdressing-v1: Customizable virtual dressing,

    F. Shen, X. Jiang, X. He, H. Ye, C. Wang, X. Du, Z. Li, and J. Tang, “Imagdressing-v1: Customizable virtual dressing,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 39, no. 7, 2025, pp. 6795– 6804

    2025

  24. [32]

    Boosting consistency in story visualization with rich-contextual conditional diffusion models,

    F. Shen, H. Ye, S. Liu, J. Zhang, C. Wang, X. Han, and Y . Wei, “Boosting consistency in story visualization with rich-contextual conditional diffusion models,” in Proceedings of the AAAI Conference on Artificial Intel- ligence, vol. 39, no. 7, 2025, pp. 6785–6794

    2025

  25. [33]

    Dinov2: Learning robust visual features without supervision,

    M. Oquab, T. Darcet, T. Moutakanni, H. V o, M. Szafraniec, V . Khalidov, P. Fernandez, D. Haziza, F. Massa, A. El-Nouby et al., “Dinov2: Learning robust visual features without supervision,” arXiv preprint arXiv:2304.07193, 2023

    2023 arXiv

  26. [34]

    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 conference on computer vision and pattern recog- nition, 2016, pp. 770–778

    2016

  27. [35]

    Searching for mobilenetv3,

    A. Howard, M. Sandler, G. Chu, L.-C. Chen, B. Chen, M. Tan, W. Wang, Y . Zhu, R. Pang, V . Vasudevan et al., “Searching for mobilenetv3,” in Proceedings of the 11 IEEE/CVF international conference on computer vision , 2019, pp. 1314–1324

    2019

  28. [36]

    Mobilevit: light-weight, general-purpose, and mobile-friendly vision transformer,

    S. Mehta and M. Rastegari, “Mobilevit: light-weight, general-purpose, and mobile-friendly vision transformer,” arXiv preprint arXiv:2110.02178 , 2021

    2021 arXiv

  29. [37]

    Fastvit: A fast hybrid vision transformer using structural reparameterization,

    P. K. A. Vasu, J. Gabriel, J. Zhu, O. Tuzel, and A. Ranjan, “Fastvit: A fast hybrid vision transformer using structural reparameterization,” in Proceedings of the IEEE/CVF international conference on computer vision , 2023, pp. 5785–5795

    2023

  30. [38]

    A convnet for the 2020s,

    Z. Liu, H. Mao, C.-Y . Wu, C. Feichtenhofer, T. Darrell, and S. Xie, “A convnet for the 2020s,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 11 976–11 986

    2022

  31. [39]

    Inception-v4, inception-resnet and the impact of residual connections on learning,

    C. Szegedy, S. Ioffe, V . Vanhoucke, and A. Alemi, “Inception-v4, inception-resnet and the impact of residual connections on learning,” in Proceedings of the AAAI conference on artificial intelligence, vol. 31, no. 1, 2017

    2017

  32. [40]

    Pyramid vision transformer: A versatile backbone for dense prediction without convo- lutions,

    W. Wang, E. Xie, X. Li, D.-P. Fan, K. Song, D. Liang, T. Lu, P. Luo, and L. Shao, “Pyramid vision transformer: A versatile backbone for dense prediction without convo- lutions,” in Proceedings of the IEEE/CVF international conference on computer vision , 2021, pp. 568–578

    2021

  33. [41]

    Augmix: A simple data pro- cessing method to improve robustness and uncertainty,

    D. Hendrycks, N. Mu, E. D. Cubuk, B. Zoph, J. Gilmer, and B. Lakshminarayanan, “Augmix: A simple data pro- cessing method to improve robustness and uncertainty,” arXiv preprint arXiv:1912.02781 , 2019

    1912 arXiv

  34. [42]

    Manifold mixup: Better representations by interpolating hidden states,

    V . Verma, A. Lamb, C. Beckham, A. Najafi, I. Mitliagkas, D. Lopez-Paz, and Y . Bengio, “Manifold mixup: Better representations by interpolating hidden states,” in International conference on machine learning . PMLR, 2019, pp. 6438–6447

    2019

  35. [43]

    mixup: Beyond empirical risk minimization,

    H. Zhang, M. Cisse, Y . N. Dauphin, and D. Lopez- Paz, “mixup: Beyond empirical risk minimization,” arXiv preprint arXiv:1710.09412, 2017

    2017 arXiv

  36. [44]

    Cutmix: Regularization strategy to train strong classi- fiers with localizable features,

    S. Yun, D. Han, S. J. Oh, S. Chun, J. Choe, and Y . Yoo, “Cutmix: Regularization strategy to train strong classi- fiers with localizable features,” in Proceedings of the IEEE/CVF international conference on computer vision , 2019, pp. 6023–6032

    2019

  37. [45]

    Pixmix: Dreamlike pictures comprehensively improve safety measures,

    D. Hendrycks, A. Zou, M. Mazeika, L. Tang, B. Li, D. Song, and J. Steinhardt, “Pixmix: Dreamlike pictures comprehensively improve safety measures,” in Proceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 16 783–16 792

    2022

  38. [46]

    Puzzle mix: Exploiting saliency and local statistics for optimal mixup,

    J.-H. Kim, W. Choo, and H. O. Song, “Puzzle mix: Exploiting saliency and local statistics for optimal mixup,” in International conference on machine learning. PMLR, 2020, pp. 5275–5285

    2020

  39. [47]

    Guidedmixup: an efficient mixup strategy guided by saliency maps,

    M. Kang and S. Kim, “Guidedmixup: an efficient mixup strategy guided by saliency maps,” in Proceedings of the AAAI conference on artificial intelligence , vol. 37, no. 1, 2023, pp. 1096–1104

    2023

  40. [48]

    Adding con- ditional control to text-to-image diffusion models,

    L. Zhang, A. Rao, and M. Agrawala, “Adding con- ditional control to text-to-image diffusion models,” in Proceedings of the IEEE/CVF international conference on computer vision , 2023, pp. 3836–3847

    2023

  41. [49]

    Ip-adapter: Text compatible image prompt adapter for text-to-image diffusion models,

    H. Ye, J. Zhang, S. Liu, X. Han, and W. Yang, “Ip-adapter: Text compatible image prompt adapter for text-to-image diffusion models,” arXiv preprint arXiv:2308.06721, 2023

    2023 arXiv

  42. [50]

    T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models,

    C. Mou, X. Wang, L. Xie, Y . Wu, J. Zhang, Z. Qi, and Y . Shan, “T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models,” in Proceedings of the AAAI conference on artificial intelligence, vol. 38, no. 5, 2024, pp. 4296–4304

    2024

  43. [2024]

    Available: https://openreview.net/forum? id=rHzapPnCgT

    [Online]. Available: https://openreview.net/forum? id=rHzapPnCgT

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.