Pith. sign in

REVIEW 3 major objections 5 minor 61 references

Adversarial Semantic Augmentation for Training Generative Adversarial Networks under Limited Data

T0 review · 3 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read This paper claims that GAN training can be augmented in semantic feature space — not image space — by perturbing features along estimated covariance directions, and that the infinite-augmentation loss has a closed-form upper bound.

desk verdict ASA-GAN transfers semantic augmentation to GAN training with broad empirical gains, but its central claim of preserving the original data distribution is asserted, not proved, and is likely a non-sequitur. read the letter →

arxiv 2502.00800 v1 pith:CK6TBYSL submitted 2025-02-02 cs.CV eess.IV

Mengping Yang , Zhe Wang , Ziqiu Chi , Dongdong Li , Wenli Du This is my paper · ORCID
classification cs.CVeess.IV
keywords generativeadversarialnetworksfew-shotimagegenerationsemanticaugmentationcovarianceestimationdatalimitedGANtraininglossupperbound
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

This paper claims that the usual trick of stretching scarce GAN training data with image transforms (flipping, cropping, color jitter) can be replaced or supplemented by perturbing the deep semantic features the discriminator extracts. The authors estimate covariance matrices of real and generated features, treat those matrices as encoding meaningful transformation directions, and sample perturbed features from a Gaussian around each original feature. They then show that the expected adversarial loss over infinitely many such samples is upper-bounded by a closed-form expression, so the augmentation can be applied implicitly by optimizing that bound rather than by explicitly sampling. The claimed payoff is better fidelity and diversity in few-shot GAN training, while leaving the original data distribution untouched and adding negligible compute. If true, this makes semantic augmentation a cheap, plug-in regularizer for any GAN.

What carries the argument

The load-bearing object is the upper bound $L_\infty$ of the expected adversarial cross-entropy loss under Gaussian feature perturbation. Starting from $F_i^*\sim\mathcal{N}(F_i, \lambda\Sigma_i)$, Jensen's inequality gives $E[\log X]\le\log E[X]$, and the Gaussian moment-generating function turns the expectation into a closed-form quadratic form $e^{(w_j-w_{\tilde y})^T F_i + (b_j-b_{\tilde y}) + \frac{\lambda}{2}(w_j-w_{\tilde y})^T\Sigma_i(w_j-w_{\tilde y})}$. This converts infinitely many augmentation samples into a single loss term. The covariance matrices $\Sigma_i$ for real and generated features are maintained online with the cited update rule, and channel and spatial attention modules improve the feature estimates that feed into the covariance computation.

What would settle it

Train the same ASA model on a dataset with a labeled attribute, extract the covariance direction corresponding to that attribute, and measure whether classifier accuracy on the attribute changes when features are perturbed along that direction. If the perturbed features do not shift the predicted attribute more than isotropic noise of the same magnitude — or if the FID of the model trained with ASA is no better than the same model with isotropic noise added to features — the semantic interpretation of the covariance directions is unsupported.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that semantic augmentation for GANs can be made implicit and infinite. Concretely, let $F_i$ be a deep feature extracted by the discriminator and let $\Sigma_i$ be the online-estimated covariance of the real or generated feature class. Sampling $F_i^* \sim \mathcal{N}(F_i, \lambda\Sigma_i)$ produces new semantic representations; the paper proves that the binary cross-entropy loss averaged over $S\to\infty$ such samples is bounded above by $L_\infty$, obtained from Jensen's inequality and the moment-generating function of a Gaussian. Optimizing $L_\infty$ therefore performs the augmentation in closed form. Because the covariance is estimated from the training features themselves and no image transform is applied, the paper argues the augmented objective still points at the original data distribution $P_d$, so the generator minimizes JS divergence with respect to $P_d$ rather than a distorted distribution. The method is claimed to be the first semantic augmentation for GANs.

Load-bearing premise

The premise that gives the argument its force is that adding Gaussian noise whose covariance matches a feature class actually moves a feature along semantically meaningful directions — background, expression, viewpoint — rather than adding arbitrary noise; if the deep feature manifold is not locally linear and approximately Gaussian, the perturbed features are not semantic, and the distribution-preservation and diversity claims no longer follow.

Editorial extensions

If this is right

  • GAN discriminators can be regularized against overfitting in low-data regimes without any image-level transform, removing the risk of leaking augmented-image artifacts into the generator's target distribution.
  • The method acts as a drop-in loss term: any GAN whose discriminator exposes deep features can swap its classification loss for $L_\infty$, so semantic augmentation should transfer across architectures without redesigning the backbone.
  • Because optimizing $L_\infty$ is equivalent to averaging over infinitely many augmentations, the usual hyperparameter $S$ (the number of augmentation samples) disappears, and training cost stays close to the unaugmented baseline.
  • Training both generator and discriminator with ASA produces larger gains than augmenting the discriminator alone, which the paper's experiments show and which implies the semantic signal also helps the generator's feature geometry.
  • The same loss still improves FID on larger datasets such as CelebA, AFHQ, and FFHQ, so the benefit is not restricted to the extreme few-shot setting.

Reading between the lines

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

  • A direct consequence the authors leave implicit: because ASA works in feature space and touches no pixels, it can be stacked on top of image-level augmentation, and the paper's own comparison figure suggests the two should be complementary rather than competing.
  • The upper-bound construction is a feature-space analogue of mixup-style regularization, since it rewards the discriminator for linear behavior around each feature; one testable prediction is that ASA should also improve robustness to small feature perturbations, not only generation FID.
  • The authors' stated limitation for imbalanced or content-diverse data suggests a concrete extension: replace the single covariance estimate with a mixture or class-conditional estimate so perturbation directions adapt to local structure, and test whether that rescues long-tailed low-data regimes.
  • If semantically meaningful directions are what matter, then estimating covariance from a pretrained feature extractor might yield larger gains than the features of a discriminator trained from scratch; the paper does not test this variant.
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

3 major / 5 minor

Summary. The paper proposes Adversarial Semantic Augmentation (ASA) for GAN training under limited data. It estimates covariance matrices of real and generated semantic features, defines an augmented feature distribution F*_i ~ N(F_i, λΣ_i), and derives an upper bound L∞ of the expected adversarial cross-entropy loss using Jensen's inequality and the Gaussian moment-generating function. By optimizing L∞, the method implicitly performs infinite semantic augmentation without explicit sampling. The authors claim that ASA makes no change to the original data distribution, introduces negligible computational overhead, and is the first semantic augmentation technique for GANs. Experiments on 20 few-shot and 3 large-scale datasets at resolutions from 256×256 to 1024×1024 report improved FID and IS over several baselines, with additional visualizations and ablations.

Significance. If the results hold, ASA is a computationally cheap and architecture-agnostic regularization that improves GAN synthesis under limited data, and the upper-bound derivation is an elegant way to avoid explicit augmentation sampling. The empirical evaluation is broad and consistently shows gains over strong baselines. However, the central theoretical claim — that ASA preserves the original data distribution — is asserted, not proven, and the current argument in Sec. 3.4 is a non-sequitur. In addition, the notation of the main loss equation (Eq. 7) is ambiguous. These issues affect the paper's main contributions as stated, but they are addressable by either providing a rigorous proof under explicit assumptions or weakening the theoretical claims. The method itself may still be valuable after revision.

major comments (3)
  1. [Sec. 3.4] The claim that ASA "makes no change to the original distribution" (abstract, contributions, and conclusion) is not proven. The argument that estimating Σ from real features and optimizing L∞ "suggests" the generator minimizes JS(P_g||P_d) is a non-sequitur. At a population optimum of the modified game, the discriminator sees the smoothed distributions P_r * N(0, λΣ_r) and P_f * N(0, λΣ_f); equality of these two convolutions does not imply equality of P_r and P_f when the smoothing kernels differ. Deconvolution is non-unique, so the generator need not recover the original data distribution. Please either supply a rigorous proof under explicit assumptions (e.g., Gaussian feature distributions with shared covariance) or remove/weaken the distribution-preservation claim.
  2. [Eq. 7] The main loss L∞ in Eq. 7 uses three weight vectors w, w_f, and w_r, whereas the derivation in Eqs. 4–6 is for a binary classification problem with C=2 classes, which would require only two weight vectors. The relationship between these notations is undefined, making the core objective ambiguous and hindering reproducibility. Please rewrite Eq. 7 in the notation of Eq. 6, or explicitly define the roles of w, w_f, and w_r and how they implement the binary real/generated classification.
  3. [Eq. 6] In the displayed line after applying Jensen's inequality, the exponent contains f*_i (the random augmented feature) rather than f_i (the original feature). The correct expression, following the moment-generating function computation, is exp((w_j−w_ỹ_i)^T f_i + (b_j−b_ỹ_i) + (λ/2)(w_j−w_ỹ_i)^T Σ_i (w_j−w_ỹ_i)). As written, the upper bound still depends on the random variable f*_i, which contradicts the claim that L∞ is a deterministic expression. The proof text later uses f_i, so this is likely a typo, but it must be corrected for the derivation to be valid.
minor comments (5)
  1. [Sec. 3.2] The text contains typos: "Jesen's inequality" should be "Jensen's inequality," and the title of Sec. 3.4, "Jesen-Shannon," should be "Jensen-Shannon."
  2. [Sec. 3.2 (proof)] In the proof after Eq. 8, the notation σ = λ(w_j−w_ỹ_i)^T Σ_i (w_j−w_ỹ_i) is incorrect; this quantity is the variance σ², not the standard deviation σ. The moment-generating function uses σ² in the exponent.
  3. [Sec. 4.1] FID results are reported as the best of three runs, which is a biased statistic. Please report the mean and standard deviation, or at least consistently report the same statistic for all methods and datasets.
  4. [Tables 3 and 4] There are minor naming inconsistencies: "templet" is likely "temple," and "moongate" should be capitalized consistently. Please proofread the dataset names.
  5. [Contributions] The claim that ASA is "the first work that employs semantic augmentation for GANs" is strong and not adequately contextualized. Prior semantic augmentation works (e.g., [29], [30]) are cited, but the discussion does not explain why they cannot be directly applied to GAN training. Please substantiate or soften this novelty claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derived L_infinity objective follows from a stated Gaussian-augmentation assumption via Jensen and the moment-generating function, and the reported FID/IS gains are held-out empirical measurements rather than quantities forced by the derivation.

full rationale

The paper's central derivation in Sec. 3.2 starts from an explicitly stated augmentation model F_i* ~ N(F_i, lambda*Sigma_i) and derives an upper bound L_infinity on the expected adversial cross-entropy loss using Jensen's inequality and the Gaussian moment-generating function. This is a mathematical consequence of the stated assumption, not a result whose conclusion is built into its premises. The covariance matrices are estimated from real and generated features, but the reported improvements are measured by FID and IS on held-out or separate evaluation sets, so no fitted parameter is renamed as a prediction. The paper does cite prior work [29] for online covariance estimation and semantic augmentation, but that citation is not by the present authors and is used as an implementation tool rather than as a uniqueness theorem or a load-bearing proof of the main claim. The self-citations ([7], [19], [20], [21]) are related-work references and do not carry the derivation. The Sec. 3.4 claim that ASA 'makes no change to the original distribution' is asserted rather than proved, and the sentence 'the consistent optimize objective suggests' is a non-sequitur; however, that is a correctness or rigor concern, not circularity, because no equation in the paper makes the original distribution equal to the augmented distribution by construction. Accordingly, no circular step can be exhibited, and the circularity score is 0.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The method rests on Gaussian and semantic-direction assumptions about deep feature spaces, a tuned lambda schedule, and the adaptation of classifier semantic augmentation to GANs. No new physical entities or external objects are introduced.

free parameters (3)
  • lambda augmentation strength = t/T (current iteration divided by total iterations)
    Hand-chosen schedule; the paper reports that a fixed lambda performs worse than the dynamic schedule, so the schedule is tuned to improve results.
  • batch size m_j = 8
    Chosen in implementation; it directly affects the online covariance estimates and the normalization of the upper-bound loss.
  • total iterations T = 100,000
    Training budget; checkpoints are saved every 10K iterations and the best FID checkpoint is selected for reporting.
assumptions (5)
  • domain assumption Semantic features of real and generated images follow a Gaussian distribution, so F_i* ~ N(F_i, lambda*Sigma_i) and the Gaussian moment-generating function can be applied.
    Used in Sec. 3.2 to derive Eq. 6; the paper does not empirically validate Gaussianity of the deep features.
  • domain assumption Covariance directions estimated from feature statistics correspond to meaningful semantic transformations such as background or expression changes.
    This is the core semantic claim of the method; the paper offers only activation-map visualizations as supporting evidence.
  • domain assumption Optimizing the upper bound L_infinity with online covariance estimates is equivalent to sampling infinitely many augmented features.
    The bound holds for a fixed covariance, but in training the covariance is updated every iteration, so the equivalence is approximate.
  • domain assumption Binary pseudo-label classification between real and fake, with separate covariances for each class, yields a valid adversarial objective for GAN training.
    Used to define Eq. 7; the paper does not formally connect this upper-bound loss to the standard GAN min-max objective.
  • standard math Jensen's inequality and the moment-generating function of a Gaussian are valid background mathematical tools.
    These tools are used correctly in the derivation of Eq. 6.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Adversarial Semantic Augmentation for Training Generative Adversarial Networks under Limited Data." pith.science (2026). https://pith.science/paper/CK6TBYSL

@misc{pith2026250200800,
  author       = {Pith},
  title        = {Pith review of: Adversarial Semantic Augmentation for Training Generative Adversarial Networks under Limited Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CK6TBYSL}},
  note         = {Machine review of arXiv:2502.00800}
}
read the original abstract

Generative adversarial networks (GANs) have made remarkable achievements in synthesizing images in recent years. Typically, training GANs requires massive data, and the performance of GANs deteriorates significantly when training data is limited. To improve the synthesis performance of GANs in low-data regimes, existing approaches use various data augmentation techniques to enlarge the training sets. However, it is identified that these augmentation techniques may leak or even alter the data distribution. To remedy this, we propose an adversarial semantic augmentation (ASA) technique to enlarge the training data at the semantic level instead of the image level. Concretely, considering semantic features usually encode informative information of images, we estimate the covariance matrices of semantic features for both real and generated images to find meaningful transformation directions. Such directions translate original features to another semantic representation, e.g., changing the backgrounds or expressions of the human face dataset. Moreover, we derive an upper bound of the expected adversarial loss. By optimizing the upper bound, our semantic augmentation is implicitly achieved. Such design avoids redundant sampling of the augmented features and introduces negligible computation overhead, making our approach computation efficient. Extensive experiments on both few-shot and large-scale datasets demonstrate that our method consistently improve the synthesis quality under various data regimes, and further visualized and analytic results suggesting satisfactory versatility of our proposed method.

Figures

Figures reproduced from arXiv: 2502.00800 by the authors.

Figure 1
Figure 1. Comparison between conventional and semantic data aug￾mentation. Conventional and semantic data augmentation perform im￾ages at the image and semantic level respectively, making them com￾plementary to each other. directions, facilitating more diverse augmented images. One straightforward way to find such directions is to define transformation directions manually, e.g., define the trans￾formation “get old”, “laugh” f… view at source ↗
Figure 2
Figure 2. Adversarial Semantic Augmentation for Training GANs under Limited Data. (left) Synthetic images of high fidelity and diversity produced by our model trained on the limited MetFace [14] dataset. These synthetic images share similar semantics but varying details such as age, gender, hairstyle, etc. (right) FID (lower is better) comparison results of our method with state-of-the-art method FastGAN [15] on 100-shot data… view at source ↗
Figure 3
Figure 3. Activation from different layers of the network on the 100- shot-Obama dataset. Different distributional information is captured at different semantic level, and the semantic features goes from general to abstract as the network goes deeper. 3.2 Adversarial Semantic Augmentation [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: The overall framework of our proposed adversarial semantic augmentation(ASA) model. We augment the training data in the semantic space by translating the features of both real and fake images along meaningful transformation directions. Such directions are obtained by e…
Figure 5
Figure 5. Figure 5: Qualitative comparison between our approach and current state-of-the-art models StyleGAN2 [4] and FastGAN [15]. The images are synthesized by the saved checkpoints with the lowest FID corresponding to each model. The red circles highlight areas with obvious artifacts a…
Figure 6
Figure 6. Figure 6: Comparison on FID dynamics of our model and DiffAug [17], ADA [14], FastGAN [15]. Our model converges faster and stabilizes the training process, alleviating the overfitting issue. We further experiment on higher resolution datasets, including 512*512 and 1024*1024. As…
Figure 7
Figure 7. Figure 7: t-SNE [58] visualization on Art Painting. 1 and 0 represent the real and synthesized images, respectively. We use the whole training images as the real distribution and randomly generate the same number of images as the training images with our generator. 4.3 More Disc…
Figure 8
Figure 8. Figure 8: Analysis of training loss with and without semantic aug￾mentation. Our semantic augmentation reduce the gap between the discriminator‘s loss and the generator‘s loss, leading to a better Nash equilibrium [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: (a) Impact of augmenting strengths. (b) Ablation Study on attention and adversarial semantic augmentation. tation. The result in [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: Nearest real images to the generated one. For each pair of images, the left images are generated our model and the right images are real images found from training sets. Zoom in for a better view [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]
Figure 11
Figure 11. Figure 11: Latent space interpolation results of our model. The smooth interpolation results indicate that our model performs satisfactory per￾formance and suffers from less overfitting. Zoom in for a better view. from overfitting. 5 LIMITATIONS AND CONCLUSION Limitations. Despi…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 52 canonical work pages

  1. [1]

    Generative adversarial nets,

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio, “Generative adversarial nets,” Advances in neural information processing systems , vol. 27, 2014

    work page 2014

  2. [2]

    Large scale gan training for high fidelity natural image synthesis,

    A. Brock, J. Donahue, and K. Simonyan, “Large scale gan training for high fidelity natural image synthesis,” in International confer- ence on machine learning , 2018

    work page 2018

  3. [3]

    Progressive growing of gans for improved quality, stability, and variation,

    T. Karras, T. Aila, S. Laine, and J. Lehtinen, “Progressive growing of gans for improved quality, stability, and variation,” in Interna- tional Conference on Learning Representations , 2018

    work page 2018

  4. [4]

    Analyzing and improving the image quality of stylegan,

    T. Karras, S. Laine, M. Aittala, J. Hellsten, J. Lehtinen, and T. Aila, “Analyzing and improving the image quality of stylegan,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 8110–8119

    2020

  5. [5]

    Alias-free generative adversarial networks,

    T. Karras, M. Aittala, S. Laine, E. H ¨ark¨onen, J. Hellsten, J. Lehti- nen, and T. Aila, “Alias-free generative adversarial networks,” Advances in Neural Information Processing Systems , vol. 34, 2021

    work page 2021

  6. [6]

    Alias-free generative adversarial networks,

    T. Karras, M. Aittala, S. Laine, E. H ¨ark¨onen, J. Hellsten, J. Lehti- nen, and T. Aila, “Alias-free generative adversarial networks,” in Advances in Neural Information Processing Systems , M. Ranzato, A. Beygelzimer, Y. Dauphin, P . Liang, and J. W. Vaughan, Eds., vol. 34, 2021

    work page 2021

  7. [7]

    Wavegan: Frequency- aware gan for high-fidelity few-shot image generation,

    M. Yang, Z. Wang, Z. Chi, and W. Feng, “Wavegan: Frequency- aware gan for high-fidelity few-shot image generation,” in Euro- pean Conference on Computer Vision , 2022, pp. 1–17

    work page 2022

  8. [8]

    Unpaired image-to- image translation using cycle-consistent adversarial networks,

    J.-Y. Zhu, T. Park, P . Isola, and A. A. Efros, “Unpaired image-to- image translation using cycle-consistent adversarial networks,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2017, pp. 2242–2251

    work page 2017

Show all 61 references
  1. [9]

    Semantic image syn- thesis with spatially-adaptive normalization,

    T. Park, M.-Y. Liu, T.-C. Wang, and J.-Y. Zhu, “Semantic image syn- thesis with spatially-adaptive normalization,” in 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2019, pp. 2337–2346

    2019

  2. [10]

    Stargan v2: Diverse image synthesis for multiple domains,

    Y. Choi, Y. Uh, J. Yoo, and J.-W. Ha, “Stargan v2: Diverse image synthesis for multiple domains,” in 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2020, pp. 8188–8197

    2020

  3. [11]

    Pepsi++: Fast and lightweight network for image inpainting,

    Y.-G. Shin, M.-C. Sagong, Y.-J. Yeo, S.-W. Kim, and S.-J. Ko, “Pepsi++: Fast and lightweight network for image inpainting,” IEEE T ransactions on Neural Networks and Learning Systems , vol. 32, no. 1, pp. 252–265, 2021

    2021

  4. [12]

    Generative image inpainting with contextual attention,

    J. Yu, Z. Lin, J. Yang, X. Shen, X. Lu, and T. S. Huang, “Generative image inpainting with contextual attention,” in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2018, pp. 5505–5514

    2018

  5. [13]

    Image inpainting via condi- tional texture and structure dual generation,

    X. Guo, H. Yang, and D. Huang, “Image inpainting via condi- tional texture and structure dual generation,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2021, pp. 14 134–14 143

    2021

  6. [14]

    Training generative adversarial networks with limited data,

    T. Karras, M. Aittala, J. Hellsten, S. Laine, J. Lehtinen, and T. Aila, “Training generative adversarial networks with limited data,” Ad- vances in Neural Information Processing Systems , vol. 33, pp. 12 104– 12 114, 2020

    2020

  7. [15]

    Towards faster and stabilized gan training for high-fidelity few-shot image synthesis,

    B. Liu, Y. Zhu, K. Song, and A. Elgammal, “Towards faster and stabilized gan training for high-fidelity few-shot image synthesis,” in International Conference on Learning Representations , 2021. 11

    2021

  8. [16]

    Spectral regularization for combating mode collapse in gans,

    K. Liu, W. Tang, F. Zhou, and G. Qiu, “Spectral regularization for combating mode collapse in gans,” in Proceedings of the IEEE international conference on computer vision , 2019, pp. 6382–6390

    2019

  9. [17]

    Differentiable augmentation for data-efficient gan training,

    S. Zhao, Z. Liu, J. Lin, J.-Y. Zhu, and S. Han, “Differentiable augmentation for data-efficient gan training,” Advances in Neural Information Processing Systems , vol. 33, pp. 7559–7570, 2020

    2020

  10. [18]

    Few-shot image generation via cross-domain corre- spondence,

    U. Ojha, Y. Li, J. Lu, A. A. Efros, Y. J. Lee, E. Shechtman, and R. Zhang, “Few-shot image generation via cross-domain corre- spondence,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 10 743–10 752

    2021

  11. [19]

    Fregan: exploiting frequency components for training gans under limited data,

    M. Yang, Z. Wang, Z. Chi, and Y. Zhang, “Fregan: exploiting frequency components for training gans under limited data,” Ad- vances in Neural Information Processing Systems , vol. 35, pp. 33 387– 33 399, 2022

    2022

  12. [20]

    Protogan: Towards high diversity and fidelity image synthesis under limited data,

    M. Yang, Z. Wang, Z. Chi, and W. Du, “Protogan: Towards high diversity and fidelity image synthesis under limited data,” Infor- mation Sciences, vol. 632, pp. 698–714, 2023

    2023

  13. [21]

    Image synthesis under limited data: A survey and taxonomy,

    M. Yang and Z. Wang, “Image synthesis under limited data: A survey and taxonomy,” arXiv preprint arXiv:2307.16879 , 2023

    2023 arXiv

  14. [22]

    Image generation from small datasets via batch statistics adaptation,

    A. Noguchi and T. Harada, “Image generation from small datasets via batch statistics adaptation,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2019, pp. 2750–2758

    2019

  15. [23]

    Freeze the discriminator: a simple baseline for fine-tuning gans,

    S. Mo, M. Cho, and J. Shin, “Freeze the discriminator: a simple baseline for fine-tuning gans,” in CVPR AI for Content Creation Workshop, 2020

    2020

  16. [24]

    Minegan: Effective knowledge transfer from gans to target domains with few images,

    Y. Wang, A. Gonzalez-Garcia, D. Berga, L. Herranz, F. S. Khan, and J. v. d. Weijer, “Minegan: Effective knowledge transfer from gans to target domains with few images,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 9332– 9341

    2020

  17. [25]

    Improved consistency regularization for gans,

    Z. Zhao, S. Singh, H. Lee, Z. Zhang, A. Odena, and H. Zhang, “Improved consistency regularization for gans,” arXiv preprint arXiv:2002.04724, 2020

    2002 arXiv

  18. [26]

    On data augmentation for gan training,

    N.-T. Tran, V .-H. Tran, N.-B. Nguyen, T.-K. Nguyen, and N.-M. Cheung, “On data augmentation for gan training,” IEEE T ransac- tions on Image Processing , vol. 30, pp. 1882–1897, 2021

    2021

  19. [27]

    Better mixing via deep representations,

    Y. Bengio, G. Mesnil, Y. Dauphin, and S. Rifai, “Better mixing via deep representations,” in International conference on machine learning, 2013, pp. 552–560

    2013

  20. [28]

    Deep feature interpolation for image content changes,

    P . Upchurch, J. Gardner, G. Pleiss, R. Pless, N. Snavely, K. Bala, and K. Weinberger, “Deep feature interpolation for image content changes,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2017, pp. 6090–6099

    2017

  21. [29]

    Regu- larizing deep networks with semantic data augmentation,

    Y. Wang, G. Huang, S. Song, X. Pan, Y. Xia, and C. Wu, “Regu- larizing deep networks with semantic data augmentation,” IEEE T ransactions on Pattern Analysis and Machine Intelligence , pp. 1–1, 2021

    2021

  22. [30]

    Transferable semantic augmentation for domain adaptation,

    S. Li, M. Xie, K. Gong, C. H. Liu, Y. Wang, and W. Li, “Transferable semantic augmentation for domain adaptation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2021, pp. 11 516–11 525

    2021

  23. [31]

    Generative adversarial networks for image and video synthesis: Algorithms and applications,

    M.-Y. Liu, X. Huang, J. Yu, T.-C. Wang, and A. Mallya, “Generative adversarial networks for image and video synthesis: Algorithms and applications,” Proceedings of the IEEE , vol. 109, no. 5, pp. 839– 862, 2021

    2021

  24. [32]

    Deep generative modelling: A comparative review of vaes, gans, nor- malizing flows, energy-based and autoregressive models,

    S. Bond-Taylor, A. Leach, Y. Long, and C. G. Willcocks, “Deep generative modelling: A comparative review of vaes, gans, nor- malizing flows, energy-based and autoregressive models,” CoRR, vol. abs/2103.04922, 2021

    2021 arXiv

  25. [33]

    Transferring gans: generating images from limited data,

    Y. Wang, C. Wu, L. Herranz, J. van de Weijer, A. Gonzalez-Garcia, and B. Raducanu, “Transferring gans: generating images from limited data,” in European Conference on Computer Vision , 2018, pp. 218–234

    2018

  26. [34]

    Minegan: effective knowledge transfer from gans to target domains with few images,

    Y. Wang, A. Gonzalez-Garcia, D. Berga, L. Herranz, F. S. Khan, and J. v. d. Weijer, “Minegan: effective knowledge transfer from gans to target domains with few images,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 9332– 9341

    2020

  27. [35]

    Minegan++: Mining generative models for efficient knowledge transfer to limited data domains,

    Y. Wang, A. Gonzalez-Garcia, C. Wu, L. Herranz, F. S. Khan, S. Jui, and J. van de Weijer, “Minegan++: Mining generative models for efficient knowledge transfer to limited data domains,” CoRR, vol. abs/2104.13742, 2021

    2021 arXiv

  28. [36]

    Few-shot image gen- eration with elastic weight consolidation,

    Y. Li, R. Zhang, J. Lu, and E. Shechtman, “Few-shot image gen- eration with elastic weight consolidation,” in Advances in Neural Information Processing Systems , vol. 33, 2020, pp. 15 885–15 896

    2020

  29. [37]

    Few-shot adaptation of generative adversarial networks,

    E. Robb, W.-S. Chu, A. Kumar, and J.-B. Huang, “Few-shot adaptation of generative adversarial networks,” arXiv preprint arXiv:2010.11943, 2020

    2010 arXiv

  30. [38]

    Reg- ularizing generative adversarial networks under limited data,

    H.-Y. Tseng, L. Jiang, C. Liu, M.-H. Yang, and W. Yang, “Reg- ularizing generative adversarial networks under limited data,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 7921–7931

    2021

  31. [39]

    Projected gans converge faster,

    A. Sauer, K. Chitta, J. M ¨uller, and A. Geiger, “Projected gans converge faster,” in Advances in Neural Information Processing Sys- tems, M. Ranzato, A. Beygelzimer, Y. Dauphin, P . Liang, and J. W. Vaughan, Eds., vol. 34, 2021

    2021

  32. [40]

    Ensem- bling off-the-shelf models for gan training,

    N. Kumari, R. Zhang, E. Shechtman, and J.-Y. Zhu, “Ensem- bling off-the-shelf models for gan training,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 10 651–10 662

    2022

  33. [41]

    mixup: Beyond empirical risk minimization,

    H. Zhang, M. Cisse, Y. N. Dauphin, and D. Lopez-Paz, “mixup: Beyond empirical risk minimization,” in International Conference on Learning Representations , 2018

    2018

  34. [42]

    A survey on image data augmentation for deep learning,

    C. Shorten and T. M. Khoshgoftaar, “A survey on image data augmentation for deep learning,” Journal of Big Data , vol. 6, no. 1, pp. 1–48, 2019

    2019

  35. [43]

    Image augmentations for gan training,

    Z. Zhao, Z. Zhang, T. Chen, S. Singh, and H. Zhang, “Image augmentations for gan training,” arXiv preprint arXiv:2006.02595 , 2020

    2006 arXiv

  36. [44]

    Consistency regularization for generative adversarial networks,

    H. Zhang, Z. Zhang, A. Odena, and H. Lee, “Consistency regularization for generative adversarial networks,” CoRR, vol. abs/1910.12027, 2019

    1910 arXiv

  37. [45]

    Training {gan}s with stronger augmenta- tions via contrastive discriminator,

    J. Jeong and J. Shin, “Training {gan}s with stronger augmenta- tions via contrastive discriminator,” in International Conference on Learning Representations, 2021

    2021

  38. [46]

    Data-efficient in- stance generation from instance discrimination,

    C. Yang, Y. Shen, Y. Xu, and B. Zhou, “Data-efficient in- stance generation from instance discrimination,” arXiv preprint arXiv:2106.04566, 2021

    2021 arXiv

  39. [47]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium,

    M. Heusel, H. Ramsauer, T. Unterthiner, B. Nessler, and S. Hochre- iter, “Gans trained by a two time-scale update rule converge to a local nash equilibrium,” in Advances in Neural Information Processing Systems, vol. 30, 2017, pp. 6626–6637

    2017

  40. [48]

    Unsupervised represen- tation learning with deep convolutional generative adversarial networks,

    A. Radford, L. Metz, and S. Chintala, “Unsupervised represen- tation learning with deep convolutional generative adversarial networks,” in International Conference on Learning Representations , 2016

    2016

  41. [49]

    Cbam: Convolutional block attention module,

    S. Woo, J. Park, J.-Y. Lee, and I. S. Kweon, “Cbam: Convolutional block attention module,” in Proceedings of the European conference on computer vision (ECCV) , 2018, pp. 3–19

    2018

  42. [50]

    Geometric gan,

    J. H. Lim and J. C. Ye, “Geometric gan,” arXiv preprint arXiv:1705.02894, 2017

    2017 arXiv

  43. [51]

    Learning hybrid image templates (hit) by information projection,

    Z. Si and S.-C. Zhu, “Learning hybrid image templates (hit) by information projection,” IEEE T ransactions on pattern analysis and machine intelligence, vol. 34, no. 7, pp. 1354–1367, 2011

    2011

  44. [52]

    Brecahad: a dataset for breast cancer histopathological annotation and diagno- sis,

    A. Aksac, D. J. Demetrick, T. Ozyer, and R. Alhajj, “Brecahad: a dataset for breast cancer histopathological annotation and diagno- sis,” BMC Research Notes , vol. 12, no. 1, p. 82, 2019

    2019

  45. [53]

    A visual vocabulary for flower classification,

    M.-E. Nilsback and A. Zisserman, “A visual vocabulary for flower classification,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , vol. 2, 2006, pp. 1447–1454

    2006

  46. [54]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium,

    M. Heusel, H. Ramsauer, T. Unterthiner, B. Nessler, and S. Hochre- iter, “Gans trained by a two time-scale update rule converge to a local nash equilibrium,” Advances in neural information processing systems, vol. 30, 2017

    2017

  47. [55]

    Improved techniques for training gans,

    T. Salimans, I. Goodfellow, W. Zaremba, V . Cheung, A. Radford, and X. Chen, “Improved techniques for training gans,” inAdvances in Neural Information Processing Systems , vol. 29, 2016, pp. 2234– 2242

    2016

  48. [56]

    Spectral nor- malization for generative adversarial networks,

    T. Miyato, T. Kataoka, M. Koyama, and Y. Yoshida, “Spectral nor- malization for generative adversarial networks,” in International Conference on Learning Representations , 2018

    2018

  49. [57]

    Adam: A method for stochastic optimiza- tion,

    D. P . Kingma and J. Ba, “Adam: A method for stochastic optimiza- tion,” arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

  50. [58]

    Visualizing data using t-sne

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

    2008

  51. [59]

    Deep learning face attributes in the wild,

    Z. Liu, P . Luo, X. Wang, and X. Tang, “Deep learning face attributes in the wild,” in Proceedings of the IEEE international conference on computer vision, 2015, pp. 3730–3738

    2015

  52. [60]

    A style-based generator archi- tecture for generative adversarial networks,

    T. Karras, S. Laine, and T. Aila, “A style-based generator archi- tecture for generative adversarial networks,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2019, pp. 4401–4410. 12

    2019

  53. [61]

    The unreasonable effectiveness of deep features as a perceptual metric,

    R. Zhang, P . Isola, A. A. Efros, E. Shechtman, and O. Wang, “The unreasonable effectiveness of deep features as a perceptual metric,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 586–595

    2018

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.