Pith. sign in

REVIEW 3 major objections 5 minor 53 references

FlexPose: Pose Distribution Adaptation with Limited Guidance

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read FlexPose claims that adapting a pre-trained pose generator to a new pose distribution requires fine-tuning only a few linear blocks of the style transform, using as few as 12 to 30 target annotations.

desk verdict A solid few-shot pose distribution adaptation method with a clear idea and consistent experiments, though the fixed layer choice is an empirical tuning decision rather than a validated principle. read the letter →

arxiv 2412.13463 v1 pith:SCRWEHZC submitted 2024-12-18 cs.CV cs.AI

Zixiao Wang , Junwu Weng , Mengyuan Liu , Bei Yu This is my paper · ORCID
classification cs.CVcs.AI
keywords posedistributionadaptationfew-shotlearningStyleGANskeletonimagegenerationdomainannotationkeypointmatchinglineartransfer
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

FlexPose sets out to show that the gap between two pose datasets is mostly a geometric transformation, not a difference in pose content. The paper treats joint annotations as skeleton images and adapts a StyleGAN that already models the source pose distribution by fine-tuning a few linear blocks of its style transform, using as few as 12 to 30 target annotations. If this works as claimed, a pre-trained pose generator can be re-targeted cheaply and then produce unlimited pose annotations for a new dataset. The paper reports cross-dataset experiments on human pose annotation, face landmark annotation, and pose-conditioned image generation where FlexPose outperforms transfer baselines.

What carries the argument

The load-bearing object is the transfer function $\tau$, a block-diagonal matrix applied to the style code at the output of the source transform $\delta_s$. Only selected blocks of $\tau$ are learned; the rest are fixed as identity, so the adaptation is sparse and linear. Pose-mixup (keypoint-space interpolation of target poses) enlarges the guidance set, and the frozen StyleGAN backbone keeps the pose prior fixed. Together these pieces localize the domain shift to a single geometric layer, which is what makes few-shot adaptation work in the paper's account.

What would settle it

Apply FlexPose to a source-target pair whose pose shift also includes viewpoint-dependent foreshortening or non-uniform body scaling, and check whether downstream keypoint accuracy still beats simply training on the source poses. If PCK or MSE stays at the no-adaptation baseline level, the block-diagonal linear-shift assumption is falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that a pose distribution can be transferred from source to target by changing only the affine style code, while freezing the generative backbone that holds the shared hinge-structure prior. Formally, the generator factorizes as $g = \phi \circ \delta$, where $\phi$ preserves the pose prior and $\delta$ maps that prior to a domain's skeleton images; FlexPose learns a block-diagonal transfer matrix $\tau$ so that the target generator is $g_t = \phi \circ (\tau \circ \delta_s)$. With $\tau$ constrained to be identity except for a small number of blocks (layer 3 in most experiments), plus interpolation-based Pose-mixup and linear/sparse regularization, this single small module is enough to align the generated pose distribution with the target in the paper's experiments. The adapted generator then yields synthetic keypoint annotations that improve unsupervised matching and downstream pose-related tasks over adapting the whole generator.

Load-bearing premise

The premise is that different pose datasets differ only in a geometric transformation, and that one small linear adjustment to a single generator layer is enough to capture it.

Editorial extensions

If this is right

  • With 30 COCO-derived guide poses, the transferred generator produces skeletons whose Fréchet distance and MMD to Human3.6M are below those of the AdaGAN, FreezeD, and LoRA baselines.
  • Using the adapted skeletons to train an unsupervised keypoint detector raises PCK to 0.585 on COCO-to-Human3.6M and 0.770 on COCO-to-Simplified-Human3.6M, well above the no-adaptation baselines.
  • The same transfer works for facial landmarks: 30 target annotations from WFLW improve the detector's MSE from 18.78 to 11.64 over the baseline.
  • Mixing multiple source datasets increases source diversity and improves downstream accuracy, while increasing the number of target shots beyond 12 brings little gain.

Reading between the lines

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

  • If the linear-shift mechanism is right, FlexPose's performance on a new source-target pair is a direct probe of how geometric the domain shift is; it should degrade exactly as appearance-level differences start to dominate.
  • The recipe should transfer to other keypoint-structured domains, such as hands or animal skeletons, wherever an affine style block can express the shift; testing that would confirm the prior-sharing assumption beyond human pose.
  • The learned transfer matrix is effectively a low-dimensional measurement of the pose shift, so its rank or singular values across dataset pairs could predict how many target annotations are actually needed.
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 paper proposes FlexPose, a method to adapt a pre-trained StyleGAN-based pose generator (trained on skeleton images rendered from 2D keypoints) to a target pose distribution using only a few annotated target poses. Adaptation is performed by inserting a block-diagonal affine transfer function τ that multiplies the style codes at a small number of layers, with linear, sparse, and keypoint-mixup regularizations to avoid collapse. The adapted generator then produces synthetic poses that are used for downstream tasks such as human pose annotation via a matching algorithm, face landmark annotation, and pose-conditional image generation. Experiments across several source/target dataset pairs show improvements over AdaGAN, FreezeD, and LoRA baselines on the human pose annotation task, with more modest results on face landmarks.

Significance. If the central claim holds, FlexPose offers a computationally efficient (under 5 minutes on a V100) and data-efficient approach to pose distribution transfer that avoids using human appearance images, which is attractive for privacy-sensitive applications. The paper provides a clear pipeline and a reasonable set of comparisons. However, the significance is limited by the strong expressivity assumption that a single linear style block at a fixed layer can capture real cross-dataset pose shifts, and by the absence of statistical validation, especially in the face landmark experiment where gains are marginal.

major comments (3)
  1. [Section 3.3, Eq. (6), and Appendix Figure 9] The method restricts the transfer to a block-diagonal affine map τ acting on a single style layer (l=3), chosen empirically on the target guidance. However, the paper's own analysis in Section 9 shows that global transformations (rotation, scale) are best fitted at layers 3–4 while local joint rotations are best fitted at layers 5–6. Real cross-dataset shifts such as 3DHP→S-H3.6M presumably contain both local and global components, yet the paper fixes l=3 without verifying that the dominant component of each real shift is global. This is a load-bearing point: if a target shift lies primarily in the local-joint regime, the single-layer τ at l=3 cannot express it, and the central claim that a limited linear layer shift suffices for few-shot pose distribution adaptation would fail for such cases. Please either (a) provide an adaptive layer-selection criterion that does not require the target annotation set, or (b) present evidence (e.g., reconstruction losses per layer for the actual dataset shifts) that the studied shifts are captured by l=3, or (c) explicitly re-scope the claim to shifts that are predominantly global.
  2. [Tables 2 and 3] The experimental comparisons report single runs with no error bars or significance tests. In the face landmark experiment (Table 3), FlexPose is best on MSE (11.64 vs 11.77 for LoRA) but not on PCK, where AdaGAN (0.785) and FreezeD (0.779) outperform FlexPose (0.766). Differences of this size are plausibly within run-to-run noise, especially since the downstream evaluation involves a separately trained KeypointGAN. Please report means and standard deviations over at least three seeds, or otherwise demonstrate that the observed improvements are statistically reliable, before claiming state-of-the-art performance.
  3. [Abstract and Table 3] The abstract and introduction state that FlexPose 'achieves state-of-the-art performance' and 'outperforms baselines by a large margin,' but Table 3 shows that on the face landmark task FlexPose does not achieve the best PCK among the compared methods. This overclaim should be corrected: either soften the abstract to say 'state-of-the-art on human pose annotation' or provide a quantitative justification for why the face landmark result is considered competitive rather than superior.
minor comments (5)
  1. [Abstract] In the abstract, 'progress' should be 'process'.
  2. [Section 3.3] The phrase 'data-starving deep learning model' should be rephrased to 'data-scarce' or 'small-sample'.
  3. [Equation (5)] Equation (5) is typeset ambiguously; the summation index and the definition of Γ should be spelled out so readers can verify the objective.
  4. [Section 4.4] The paper says 'we set l=3 in all experiments,' yet the same section's ablation table varies l (rows #2–#4). Rephrase to clarify that l=3 is the default choice and variants are part of the ablation.
  5. [Reproducibility] A statement on code and data availability would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: FlexPose is an explicit few-shot fitting procedure whose generalization is measured on held-out target data.

full rationale

FlexPose's derivation is an explicit fitting procedure rather than a claimed derivation from first principles. The transfer function tau is trained by minimizing the perceptual reconstruction loss in Eq. (5) against the few target guide poses, and the paper's claims are then evaluated against the full target distribution (FD and MMD over all Human3.6M samples in Table 1) and on held-out downstream annotation tasks (Tables 2 and 3). The fitted quantity (reconstruction of the few-shot set) is not renamed as the predicted quantity; instead, generalization is measured on target data not used in the fit. The load-bearing premise that a pose shift can be expressed by a block-diagonal affine style perturbation (Eq. 6, with l=3) is an explicit ansatz motivated by StyleGAN's style-control properties, not a result derived from the conclusion. The paper's own Appendix (Fig. 9) shows that global versus local transformations favor different layers, which is a correctness or expressivity risk rather than a circularity. The only self-citation (Wang et al. 2023) appears in a related-work list for weakly-supervised methods and does not support any load-bearing step. No equation or evaluation metric reduces by construction to its input; the adaptation objective and the evaluation protocol are distinct.

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

No new particles, forces, dimensions, or analogous entities are introduced. The transfer matrix tau is a learned parameter, not a postulated entity. The central claim rests on two domain assumptions (the geometric-transfer prior and the validity of the source-trained beta regressor) and one empirically chosen hyperparameter (the transfer layer index).

free parameters (3)
  • Transfer layer index l = 3
    Chosen by lowest reconstruction loss in Figure 4 and Section 4.4; performance varies with this choice (Table 4 rows 1 to 4).
  • Sparse block selection in tau (Eq. 6) = only layer 3 blocks adjusted
    The block-diagonal structure of tau and which layers to adjust are set empirically, not derived from a principle.
  • Pose-mixup ratio lambda = lambda in [0,1], sampling distribution unspecified
    Used to build the extended annotation set T* in Eq. 7; no sensitivity analysis of the lambda sampling distribution is provided.
assumptions (3)
  • domain assumption Pose distributions share a common hinge-structure prior and differ only by geometric transformations such as pivot orientation, joint rotation, and bone length ratio.
    Stated in Section 3.2 as 'We assume that the pose distributions of different datasets share similar pose prior, and their distributions can transfer to one another by geometric transformations.' It motivates the whole factorization but is not justified beyond visualization.
  • ad hoc to paper StyleGAN style layers are hierarchically disentangled for skeleton images, so adjusting a single coarse layer can represent the global pose transformation.
    Assumed from natural-image StyleGAN behavior in Section 3.3 and Section 9; the layer choice is validated empirically on the target reconstruction loss.
  • domain assumption The source-trained keypoint regressor beta generalizes to target-generated skeleton images without retraining.
    In Section 3.4, beta is pre-trained on the source via Eq. 9 and used to denoise target skeleton images; cross-domain validity is assumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FlexPose: Pose Distribution Adaptation with Limited Guidance." pith.science (2026). https://pith.science/paper/SCRWEHZC

@misc{pith2026241213463,
  author       = {Pith},
  title        = {Pith review of: FlexPose: Pose Distribution Adaptation with Limited Guidance},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SCRWEHZC}},
  note         = {Machine review of arXiv:2412.13463}
}
read the original abstract

Numerous well-annotated human key-point datasets are publicly available to date. However, annotating human poses for newly collected images is still a costly and time-consuming progress. Pose distributions from different datasets share similar pose hinge-structure priors with different geometric transformations, such as pivot orientation, joint rotation, and bone length ratio. The difference between Pose distributions is essentially the difference between the transformation distributions. Inspired by this fact, we propose a method to calibrate a pre-trained pose generator in which the pose prior has already been learned to an adapted one following a new pose distribution. We treat the representation of human pose joint coordinates as skeleton image and transfer a pre-trained pose annotation generator with only a few annotation guidance. By fine-tuning a limited number of linear layers that closely related to the pose transformation, the adapted generator is able to produce any number of pose annotations that are similar to the target poses. We evaluate our proposed method, FlexPose, on several cross-dataset settings both qualitatively and quantitatively, which demonstrates that our approach achieves state-of-the-art performance compared to the existing generative-model-based transfer learning methods when given limited annotation guidance.

Figures

Figures reproduced from arXiv: 2412.13463 by the authors.

Figure 1
Figure 1. The illustration of how poses can be adapted be [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. An illustration of the FlexPose framework for pose distribution adaptation. There are three main steps in our framework: [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. An illustration of the generator decomposition. We [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (5 more)
Figure 5
Figure 5. Figure 5: Visualization of pose adapta￾tion. The left and middle of each row are generated from the same random noise. The middle aims to mimic the right [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 7
Figure 7. Figure 7: Visualization of human face landmark annotation [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: Application on pose-conditional human image gen [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: Reconstruction loss with different augmentation [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: Illustration of augmentations. All of them generate skeleton images that less frequently appear in source targets. [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 37 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Andriluka, M.; Roth, S.; and Schiele, B. 2009. Pictorial structures revisited: People detection and articulated pose estimation. In CVPR, 1014--1021. IEEE

    work page 2009

  4. [4]

    Bai, Y.; and Wang, W. 2019. Acpnet: anchor-center based person network for human pose estimation and instance segmentation. In ICME, 1072--1077. IEEE

    work page 2019

  5. [5]

    Belagiannis, V.; and Zisserman, A. 2017. Recurrent human pose estimation. In FG, 468--475. IEEE

    work page 2017

  6. [6]

    Buolamwini, J.; and Gebru, T. 2018. Gender shades: Intersectional accuracy disparities in commercial gender classification. In Conference on fairness, accountability and transparency, 77--91. PMLR

    2018

  7. [7]

    Cao, J.; Tang, H.; Fang, H.-S.; Shen, X.; Lu, C.; and Tai, Y.-W. 2019. Cross-domain adaptation for animal pose estimation. In ICCV, 9498--9507

    work page 2019

  8. [8]

    Csurka, G. 2017. Domain adaptation for visual applications: A comprehensive survey. arXiv preprint arXiv:1702.05374

    arXiv 2017

Show all 53 references
  1. [9]

    Cui, Q.; Sun, H.; Lu, J.; Li, W.; Li, B.; Yi, H.; and Wang, H. 2023. Test-time Personalizable Forecasting of 3D Human Poses. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 274--283

    2023

  2. [10]

    Daum \'e III, H. 2009. Frustratingly easy domain adaptation. arXiv preprint arXiv:0907.1815

    2009 arXiv

  3. [11]

    Ganin, Y.; and Lempitsky, V. 2015. Unsupervised domain adaptation by backpropagation. In ICML, 1180--1189. PMLR

    2015

  4. [12]

    Gecer, B.; Ploumpis, S.; Kotsia, I.; and Zafeiriou, S. 2019. Ganfit: Generative adversarial network fitting for high fidelity 3d face reconstruction. In CVPR, 1155--1164

    2019

  5. [13]

    Geng, Z.; Cao, C.; and Tulyakov, S. 2019. 3d guided fine-grained face manipulation. In CVPR, 9821--9830

    2019

  6. [14]

    J.; Pouget-Abadie, J.; Mirza, M.; Xu, B.; Warde-Farley, D.; Ozair, S.; Courville, A

    Goodfellow, I. J.; Pouget-Abadie, J.; Mirza, M.; Xu, B.; Warde-Farley, D.; Ozair, S.; Courville, A. C.; and Bengio, Y. 2014. Generative Adversarial Nets. In NeurIPS

    2014

  7. [15]

    He, K.; Gkioxari, G.; Doll \'a r, P.; and Girshick, R. 2017. Mask r-cnn. In ICCV, 2961--2969

    2017

  8. [16]

    Heusel, M.; Ramsauer, H.; Unterthiner, T.; Nessler, B.; and Hochreiter, S. 2017. Gans trained by a two time-scale update rule converge to a local nash equilibrium. NeurIPS, 30

    2017

  9. [17]

    Ho, J.; Jain, A.; and Abbeel, P. 2020. Denoising diffusion probabilistic models. NeurIPS, 33: 6840--6851

    2020

  10. [18]

    J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W

    Hu, E. J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W. 2021. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685

    2021 arXiv

  11. [19]

    Hu, S.; Sun, H.; Li, B.; Wei, D.; Li, W.; and Lu, J. 2024. Fast Adaptation for Human Pose Estimation via Meta-Optimization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 1792--1801

    2024

  12. [20]

    Inoue, N.; Furuta, R.; Yamasaki, T.; and Aizawa, K. 2018. Cross-domain weakly-supervised object detection through progressive domain adaptation. In CVPR, 5001--5009

    2018

  13. [21]

    Ionescu, C.; Papava, D.; Olaru, V.; and Sminchisescu, C. 2014. Human3.6M: Large Scale Datasets and Predictive Methods for 3D Human Sensing in Natural Environments. TPAMI, 36(7): 1325--1339

    2014

  14. [22]

    Jakab, T.; Gupta, A.; Bilen, H.; and Vedaldi, A. 2018. Unsupervised learning of object landmarks through conditional image generation. NeurIPS, 31

    2018

  15. [23]

    Jakab, T.; Gupta, A.; Bilen, H.; and Vedaldi, A. 2020. Self-supervised learning of interpretable keypoints from unlabelled videos. In CVPR, 8787--8797

    2020

  16. [24]

    J.; Jacobs, D

    Kanazawa, A.; Black, M. J.; Jacobs, D. W.; and Malik, J. 2018. End-to-end recovery of human shape and pose. In CVPR, 7122--7131

    2018

  17. [25]

    Karras, T.; Laine, S.; and Aila, T. 2019. A style-based generator architecture for generative adversarial networks. In CVPR, 4401--4410

    2019

  18. [26]

    P.; and Welling, M

    Kingma, D. P.; and Welling, M. 2014. Auto-encoding variational bayes. ICLR

    2014

  19. [27]

    B.; and Wang, X

    Li, Y.; Hao, M.; Di, Z.; Gundavarapu, N. B.; and Wang, X. 2021. Test-time personalization with a transformer for human pose estimation. Advances in Neural Information Processing Systems, 34: 2583--2597

    2021

  20. [28]

    Lin, T.-Y.; Maire, M.; Belongie, S.; Hays, J.; Perona, P.; Ramanan, D.; Doll \'a r, P.; and Zitnick, C. L. 2014. Microsoft coco: Common objects in context. In ECCV, 740--755. Springer

    2014

  21. [29]

    Long, M.; Cao, Y.; Wang, J.; and Jordan, M. 2015. Learning transferable features with deep adaptation networks. In ICML, 97--105. PMLR

    2015

  22. [30]

    Lorenz, D.; Bereska, L.; Milbich, T.; and Ommer, B. 2019. Unsupervised part-based disentangling of object shape and appearance. In CVPR, 10955--10964

    2019

  23. [31]

    Luo, Y.; Zheng, L.; Guan, T.; Yu, J.; and Yang, Y. 2019. Taking a closer look at domain shift: Category-level adversaries for semantics consistent domain adaptation. In CVPR, 2507--2516

    2019

  24. [32]

    Mehta, D.; Rhodin, H.; Casas, D.; Fua, P.; Sotnychenko, O.; Xu, W.; and Theobalt, C. 2017. Monocular 3d human pose estimation in the wild using improved cnn supervision. In 2017 international conference on 3D vision (3DV), 506--516. IEEE

    2017

  25. [33]

    Mo, S.; Cho, M.; and Shin, J. 2020. Freeze the discriminator: a simple baseline for fine-tuning gans. CVPR Workshop

    2020

  26. [34]

    Mou, C.; Wang, X.; Xie, L.; Wu, Y.; Zhang, J.; Qi, Z.; and Shan, Y. 2024. 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, volume 38, 4296--4304

    2024

  27. [35]

    Noguchi, A.; and Harada, T. 2019. Image generation from small datasets via batch statistics adaptation. In ICCV, 2750--2758

    2019

  28. [36]

    Oquab, M.; Bottou, L.; Laptev, I.; and Sivic, J. 2014. Learning and transferring mid-level image representations using convolutional neural networks. In CVPR, 1717--1724

    2014

  29. [37]

    Sagonas, C.; Tzimiropoulos, G.; Zafeiriou, S.; and Pantic, M. 2013. 300 faces in-the-wild challenge: The first facial landmark localization challenge. In ICCVW, 397--403

    2013

  30. [38]

    Shin, S.; Hwang, K.; and Sung, W. 2016. Generative knowledge transfer for neural language models. arXiv preprint arXiv:1608.04077

    2016 arXiv

  31. [39]

    A.; Samaras, D.; Paragios, N.; and Kokkinos, I

    Shu, Z.; Sahasrabudhe, M.; Guler, R. A.; Samaras, D.; Paragios, N.; and Kokkinos, I. 2018. Deforming autoencoders: Unsupervised disentangling of shape and appearance. In ECCV, 650--665

    2018

  32. [40]

    Sohl-Dickstein, J.; Weiss, E.; Maheswaranathan, N.; and Ganguli, S. 2015. Deep unsupervised learning using nonequilibrium thermodynamics. In ICML, 2256--2265. PMLR

    2015

  33. [41]

    Thewlis, J.; Bilen, H.; and Vedaldi, A. 2017. Unsupervised learning of object landmarks by factorized spatial embeddings. In ICCV, 5916--5925

    2017

  34. [42]

    Vahdat, A.; and Kautz, J. 2020. NVAE: A deep hierarchical variational autoencoder. NeurIPS, 33: 19667--19679

    2020

  35. [43]

    J.; Laptev, I.; and Schmid, C

    Varol, G.; Romero, J.; Martin, X.; Mahmood, N.; Black, M. J.; Laptev, I.; and Schmid, C. 2017. Learning from Synthetic Humans. In CVPR

    2017

  36. [44]

    Wang, D.; and Zhang, S. 2022. Contextual Instance Decoupling for Robust Multi-Person Pose Estimation. In CVPR, 11060--11068

    2022

  37. [45]

    Wang, Y.; Wu, C.; Herranz, L.; van de Weijer, J.; Gonzalez-Garcia, A.; and Raducanu, B. 2018. Transferring gans: generating images from limited data. In ECCV, 218--234

    2018

  38. [46]

    Wang, Z.; Weng, J.; Yuan, C.; and Wang, J. 2023. Truncate-split-contrast: a framework for learning from mislabeled videos. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, 2751--2758

    2023

  39. [47]

    Wu, W.; Qian, C.; Yang, S.; Wang, Q.; Cai, Y.; and Zhou, Q. 2018. Look at Boundary: A Boundary-Aware Face Alignment Algorithm. In CVPR

    2018

  40. [48]

    Yin, F.; Zhang, Y.; Cun, X.; Cao, M.; Fan, Y.; Wang, X.; Bai, Q.; Wu, B.; Wang, J.; and Yang, Y. 2022. Styleheat: One-shot high-resolution editable talking face generation via pretrained stylegan. arXiv preprint arXiv:2203.04036

    2022 arXiv

  41. [49]

    N.; and Lopez-Paz, D

    Zhang, H.; Cisse, M.; Dauphin, Y. N.; and Lopez-Paz, D. 2017. mixup: Beyond empirical risk minimization. ICLR

    2017

  42. [50]

    Zhang, L.; and Agrawala, M. 2023. Adding Conditional Control to Text-to-Image Diffusion Models. arXiv:2302.05543

    2023 arXiv

  43. [51]

    Zhang, Y.; Guo, Y.; Jin, Y.; Luo, Y.; He, Z.; and Lee, H. 2018. Unsupervised discovery of object landmarks as structural representations. In CVPR, 2694--2703

    2018

  44. [52]

    a henb \

    Zhu, J.-Y.; Kr \"a henb \"u hl, P.; Shechtman, E.; and Efros, A. A. 2016. Generative visual manipulation on the natural image manifold. In ECCV, 597--613. Springer

    2016

  45. [53]

    Zhu, J.-Y.; Park, T.; Isola, P.; and Efros, A. A. 2017. Unpaired image-to-image translation using cycle-consistent adversarial networks. In CVPR, 2223--2232

    2017

Pith tools

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