Pith. sign in

REVIEW 3 major objections 4 minor 31 references

Fine-grained Text to Image Synthesis

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read This paper proposes FG-RAT GAN, which adds an auxiliary classifier and contrastive learning with cross-batch memory to RAT GAN, and reports the lowest FID scores on CUB-200-2011 and Oxford-102.

desk verdict A modest GAN-era fine-grained text-to-image extension whose FID gains are plausible but unverified, with a likely sign error in the contrastive loss equations. read the letter →

arxiv 2412.07196 v2 pith:US5Q5QOD submitted 2024-12-10 cs.CV

classification cs.CV
keywords fine-grainedtext-to-imagesynthesisauxiliaryclassifiercontrastivelearningcross-batchmemorygenerativeadversarialnetworkCUB-200-2011Oxford-102FréchetInceptionDistance
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 argues that the RAT GAN text-to-image model can be made to respect fine-grained class distinctions by adding two inexpensive components: an auxiliary classifier that forces the discriminator to predict the image's subclass, and a contrastive loss, backed by a cross-batch memory, that pulls same-class images together and pushes different-class images apart. On the CUB-200-2011 bird dataset and Oxford-102 flower dataset the augmented model, called FG-RAT GAN, reports the lowest Fréchet Inception Distance among LAFITE, VQ-Diffusion, and RAT GAN, meaning its images are statistically closer to real photos under the Inception feature metric. The paper treats this as evidence that fine-grained label information can be leveraged efficiently to improve synthesis quality.

What carries the argument

The mechanism is an auxiliary classifier plus contrastive embedding added to the RAT GAN discriminator. The auxiliary classifier computes a categorical cross-entropy loss that grades both real and fake images against subclass labels, feeding a gradient to the generator; the contrastive branch L2-normalizes 256-D embeddings and applies a margin-based cosine similarity loss against a cross-batch memory queue of real-image embeddings, so that same-class pairs are pulled together and different-class pairs are pushed apart past margin α. The two losses are summed with the adversarial loss after an initial phase that stabilizes the embeddings through classification alone.

What would settle it

Train a version of RAT GAN that receives the same subclass labels as an additional input to the generator or discriminator but omits the auxiliary classifier and contrastive loss; if its FID matches the reported 8.66 on CUB-200-2011, the paper's attribution of the gain to the new losses is falsified. Repeating each configuration across multiple random seeds to obtain FID error bars would also test whether the reported gaps exceed run-to-run noise.

Watch

Extended reading notes

Core claim

FG-RAT GAN extends RAT GAN's discriminator with a fully connected classification head trained with categorical cross-entropy on both real and fake images, and an embedding branch trained with a contrastive loss that compares current-batch features against a queue of past real-image features. The two losses join the original adversarial loss in both generator and discriminator updates. On CUB-200-2011 the model reports FID 8.66 versus 12.12 for RAT GAN, 10.32 for VQ-Diffusion, and 10.48 for LAFITE; on Oxford-102 it reports FID 9.14 versus 12.90 for RAT GAN. The paper's claim is that these gains come from the auxiliary classifier and contrastive learning making the generator produce images that are more internally consistent within a subclass and better separated across subclasses.

Load-bearing premise

The comparison assumes that feeding subclass labels during training is not itself the cause of the FID improvement, so the gains can be attributed to the auxiliary classifier and contrastive learning rather than to the extra label signal.

Editorial extensions

If this is right

  • On both CUB-200-2011 and Oxford-102, FG-RAT GAN reports lower FID than RAT GAN, LAFITE, and VQ-Diffusion, so the added components transfer across two datasets with different class counts.
  • The full model beats both single-component ablations on FID, supporting the paper's claim that the classifier and contrastive learning are complementary rather than redundant.
  • Because only two fully connected layers (roughly 17M parameters) are added at training time, the method offers a parameter-efficient path to fine-grained text-to-image synthesis without scaling to billion-parameter autoregressive or diffusion models.
  • The generator's within-class similarity increases while cross-class similarity decreases, which is the geometric condition the contrastive loss directly enforces.

Reading between the lines

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

  • A natural stress test would run the same ablation with class labels removed from the auxiliary head entirely to see how much of the FID drop is attributable to label availability rather than to the contrastive geometry.
  • The cross-batch memory could be repurposed as a hard-negative mining mechanism for other conditional generators, not just GANs, by swapping the memory queue into diffusion or autoregressive training loops.
  • If label noise is introduced during training, the margin-based contrastive loss may degrade more gracefully than the hard classifier, a hypothesis the current single-run FID table does not test.
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 / 4 minor

Summary. The paper proposes FG-RAT GAN, an extension of RAT GAN for fine-grained text-to-image synthesis. It adds an auxiliary classifier to the discriminator and a contrastive learning loss with a cross-batch memory (XBM) mechanism, and evaluates the resulting model on CUB-200-2011 and Oxford-102. The central claim, stated in the abstract and in Section 4.5, is that FG-RAT GAN achieves the lowest FID among the compared methods (LAFITE, VQ-Diffusion, RAT GAN) on both datasets while adding only a modest number of parameters.

Significance. If the reported results are valid, the paper would offer a lightweight and conceptually simple improvement over RAT GAN, with a parameter overhead of about 17M and a claimed FID reduction from 12.12 to 8.66 on CUB-200-2011 and from 12.90 to 9.14 on Oxford-102. The use of FID as the primary metric, with an explicit discussion of the limitations of Inception Score, is a reasonable methodological choice. However, the manuscript provides no code, no statistical uncertainty quantification, and the main comparison is not controlled for the extra class-label supervision used by the proposed method; these issues prevent the central claim from being verified as written. The auxiliary-classifier idea is well motivated, and the overall approach is easy to follow, but the load-bearing evaluation and the sign of the contrastive-loss term need correction before the contribution can be assessed.

major comments (3)
  1. [Section 3.2, Eqs. (3)-(4)] The negative-pair term in the contrastive loss has the wrong sign relative to the stated objective. For a negative pair with cos_sim > alpha, the term is written as -max((cos_sim - alpha), 0), which is negative and decreases as cos_sim increases. Minimizing this term therefore encourages embeddings from different subclasses to become more similar, contradicting the text's claim that the loss 'minimizes the similarity between feature embeddings of ... different subclasses.' The correct hinge form should be +max((cos_sim - alpha), 0). Since the contrastive loss is one of the two named contributions and Table 2 attributes part of the FID improvement to it, this is not a cosmetic typo; as printed, the equations do not implement the described mechanism. The absence of released code makes it impossible to determine whether the implementation used the printed sign or a corrected one.
  2. [Section 4.5, Table 1] The comparison against LAFITE and VQ-Diffusion is not controlled for the additional supervision used by FG-RAT GAN. The proposed method trains an auxiliary classifier with ground-truth class labels, while the two baselines are label-free text-conditioned models. The sentence 'Even though we use labels during the training, label information is not an unfair advantage but a distinct characteristic of our model' is an assertion, not a controlled experiment. The ablation in Table 2 shows that RAT GAN + classifier, which also uses labels, already reduces FID from 12.12 to 9.90 on CUB, so the label signal alone explains a substantial part of the gain. A comparison with a label-conditioned baseline (e.g., ACGAN or TAC-GAN) or a version of the proposed model trained without the classifier is necessary to attribute the reported improvement to the proposed mechanisms.
  3. [Section 4.5, Tables 1 and 2] All reported IS and FID values appear to come from a single training run, with no error bars, confidence intervals, or significance tests. The claimed gains are sometimes small (e.g., IS 5.08 vs 4.99 in Table 2 on CUB, and IS 3.66 vs 3.62 on Oxford), and FID is known to vary across seeds and evaluation batches. Without multiple runs and standard deviations, the reader cannot assess whether the reported FID differences, such as 8.66 vs 9.10 in Table 2, are statistically meaningful. The paper should report mean and standard deviation over at least three seeds for each configuration.
minor comments (4)
  1. [Throughout] There are numerous typos and formatting errors, including 'Frenchet Inception Distance' (Section 4.2), 'horizontaly' (Section 4.1), 'orginal' (Section 4.4), 'auxliary' and 'discrminator' (Figure 1 caption), 'contrastive' misspelled as 'contrtastive' (Table 2), and a stray '0' at the end of Section 4.4. The manuscript should be carefully proofread.
  2. [References, [17]-[18]] Reference [17] is the cross-batch memory (XBM) source, but reference [18] is a fine-grained image recognition paper by the current authors. Citing [18] together with [17] for the XBM mechanism is misleading unless the connection is explicitly explained; the citation should be corrected or removed.
  3. [Section 5, Conclusion] The conclusion states that 'further evaluations on broader text-to-image synthesis benchmarks and more varied datasets are necessary to confirm the generalizability of our approach.' This is a self-acknowledged limitation, but the paper currently presents the method as achieving state-of-the-art performance; the scope of the claim should be aligned with this admitted limitation in the abstract and Section 4.5.
  4. [Section 4.4 and Appendix 6.1] The qualitative comparisons against DALL-E 2 and Stable Diffusion in the appendix are based only on a few cherry-picked examples and are not accompanied by any quantitative evaluation or human-study protocol. The text 'we demonstrate that our FG-RAT GAN can reach better visualized results compared with DALLE-2 and Stable Diffusion' overstates what such figures can support.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central FID claim is benchmarked externally and the proposed losses are not fitted to the metric.

full rationale

The paper's central claim is empirical: FG-RAT GAN obtains lower FID than LAFITE, VQ-Diffusion, and RAT GAN on CUB-200-2011 and Oxford-102 (Section 4.5, Table 1). These numbers are evaluated against external models with separately reported results, and FID/IS compare generated images to real-image feature statistics; neither metric is a re-packaged training loss or a fitted value from the model's own objective. The proposed components (auxiliary classifier and contrastive loss) are defined as training losses (Eqs. 1-4, 7-8), and the ablation in Table 2 attributes the FID change to those components. The loss definitions do not reduce by construction to the reported FID scores. Self-citations in Section 2 ([3], [18], [25], [26]) are related-work attributions and are not load-bearing for the main result; the backbone RAT GAN [7] and the XBM mechanism [17] are external works. The paper itself flags its dependence on fine-grained labels in the conclusion and asserts in Section 4.5 that label use is 'not an unfair advantage but a distinct characteristic of our model'; this is a potential experimental-design confound, but it is a soundness concern, not circularity. Likewise, the apparent sign inconsistency in the negative-pair term of Eqs. (3)-(4) is a correctness issue, not a derivation that assumes its own conclusion. No circular step is exhibited.

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

The central empirical claim depends on several unreported hyperparameters (margin alpha, memory size M, batch size, epoch schedule) and on the assumption that label supervision is not the source of the FID gain. No code or trained models are provided.

free parameters (5)
  • Contrastive loss margin alpha = Not reported
    Alpha is defined in Equations (3) and (4) as a margin for negative contrastive pairs but its value is never given in Section 4.3.
  • XBM memory size M = Not reported
    M controls the number of stored real-image embeddings in the cross-batch memory and is not specified.
  • Batch size N = Not reported
    Batch size appears in the contrastive loss and in training but is not reported.
  • Epoch at which contrastive loss is enabled = After several training epochs
    Section 3.3 says the contrastive loss is added 'after several training epochs' without giving the specific epoch.
  • Loss weights for L_ce and L_cl = Implicitly 1.0
    Equations (7) and (8) sum adversarial, cross-entropy, and contrastive losses with unit weights; no weighting schedule or tuning is reported.
assumptions (3)
  • domain assumption Class labels are available for all training images and are used in both discriminator and generator losses.
    The method uses fine-grained labels for 200 bird classes and 102 flower classes; the conclusion acknowledges dependence on labels as a limitation.
  • domain assumption FID computed on a single run is a reliable and stable measure of fine-grained synthesis quality.
    The paper uses FID as its primary superiority metric but reports no error bars, seeds, or statistical tests.
  • domain assumption Cosine similarity in the learned 256-dimensional embedding reflects fine-grained subclass similarity.
    The contrastive losses in Equations (3) and (4) presuppose that cosine distance in this embedding space corresponds to semantic subclass distance.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fine-grained Text to Image Synthesis." pith.science (2026). https://pith.science/paper/US5Q5QOD

@misc{pith2026241207196,
  author       = {Pith},
  title        = {Pith review of: Fine-grained Text to Image Synthesis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/US5Q5QOD}},
  note         = {Machine review of arXiv:2412.07196}
}
read the original abstract

Fine-grained text to image synthesis involves generating images from texts that belong to different categories. In contrast to general text to image synthesis, in fine-grained synthesis there is high similarity between images of different subclasses, and there may be linguistic discrepancy among texts describing the same image. Recent Generative Adversarial Networks (GAN), such as the Recurrent Affine Transformation (RAT) GAN model, are able to synthesize clear and realistic images from texts. However, GAN models ignore fine-grained level information. In this paper we propose an approach that incorporates an auxiliary classifier in the discriminator and a contrastive learning method to improve the accuracy of fine-grained details in images synthesized by RAT GAN. The auxiliary classifier helps the discriminator classify the class of images, and helps the generator synthesize more accurate fine-grained images. The contrastive learning method minimizes the similarity between images from different subclasses and maximizes the similarity between images from the same subclass. We evaluate on several state-of-the-art methods on the commonly used CUB-200-2011 bird dataset and Oxford-102 flower dataset, and demonstrated superior performance.

Figures

Figures reproduced from arXiv: 2412.07196 by the authors.

Figure 1
Figure 1. The original discrminator in Figure (a) computes GAN loss. The discrimina￾tor with auxliary classifier in Figure (b) computes categorical cross entropy loss. The discrminator with contrastive learning in Figure (c) computes contrastive learning loss. 64x1024 to 256. We then add a Softmax activation function to classify the feature into one of the predefined categories. The structure of the modified discriminator is … view at source ↗
Figure 2
Figure 2. The structure of the discriminator with auxiliary classifier and contrastive learning. The original output of the discriminator is still used to compute the GAN loss, and meanwhile followed by one fully connected layer to decrease the feature dimension. Next, the fully connected layer is followed by one embedding layer for contrastive learning. Then, the embedding layer is followed by a classifier for image classifi… view at source ↗
Figure 3
Figure 3. Examples of generated images using RAT GAN and the proposed FG-RAT GAN on the CUB bird dataset. Each row represents a different sample (image size = 256x256) and with the corresponding caption below.The first column is image class and name. The second column is the corresponding target image. The rest of other columns are the generated images from LAFITE, VQ-Diffusion, RAT GAN, and our FG-RAT GAN. As we can see, our… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Examples of generated images using RAT GAN and the proposed FG-RAT GAN with classifier and contrastive learning trained on the Oxford flower dataset. Each row represents a different sample (image size=256x256). The first column is the sample detail including class and …
Figure 5
Figure 5. Figure 5: Examples of generated images using DALLE-2, Stable Diffusion, and the pro￾posed FG-RAT GAN trained on the CUB bird dataset. Each row represents a different sample (image size=256x256). The first column is the sample detail including class and specific image name. The s…
Figure 6
Figure 6. Figure 6: Examples of generated images using DALLE-2, Stable Diffusion, and the pro￾posed FG-RAT GAN trained on the Oxford flower dataset. Each row represents a different sample (image size=256x256). The first column is the sample detail including class and specific image name. …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 24 canonical work pages

  1. [18]

    Ouyang, Y

    X. Ouyang, Y. Chen, K. Zhu and G. Agam: SwinTransFuse: Fusing swin and multi- scale transformers for fine-grained image recognition and retrieval, 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), New Orleans, LA, USA. (2022)

  2. [1]

    In: Proceedings of the 27th International Conference on Neural Information Processing Systems (NIPS’14), pp

    Goodfellow, I.J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., Bengio, Y.: Generative adversarial nets. In: Proceedings of the 27th International Conference on Neural Information Processing Systems (NIPS’14), pp. 2672–2680. MIT Press, Cambridge, MA, USA (2014)

  3. [2]

    In: Balcan, M.F., Weinberger, K.Q

    Reed, S., Akata, Z., Yan, X., Logeswaran, L., Schiele, B., Lee, H.: Generative Ad- versarial Text to Image Synthesis. In: Balcan, M.F., Weinberger, K.Q. (eds.) Pro- ceedings of The 33rd International Conference on Machine Learning, vol. 48, pp. 1060–1069. PMLR, New York, New York, USA (2016)

  4. [3]

    In: 2018 24th International Conference on Pattern Recognition (ICPR), pp

    Ouyang, X., Zhang, X., Ma, D., Agam, G.: Generating Image Sequence from De- scription with LSTM Conditional GAN. In: 2018 24th International Conference on Pattern Recognition (ICPR), pp. 2456–2461. IEEE, Beijing, China (2018)

  5. [4]

    In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recog- nition (CVPR), pp

    Xu, T., Zhang, P., Huang, Q., Zhang, H., Gan, Z., Huang, X., He, X.: AttnGAN: Fine-Grained Text to Image Generation with Attentional Generative Adversarial Networks. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recog- nition (CVPR), pp. 1316–1324. IEEE, Salt Lake City, UT, USA (2018)

  6. [5]

    In: 2022 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pp

    Zhou, Y., Chen, H., Zhang, W., Sun, Z., He, X., Fan, Y.: Towards Language-Free Training for Text-to-Image Generation. In: 2022 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pp. 17886–17896. IEEE, New Or- leans, LA, USA (2022)

  7. [6]

    In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Karras, T., Laine, S., Aittala, M., Hellsten, J., Lehtinen, J., Aila, T.: Analyzing and Improving the Image Quality of StyleGAN. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 8107–8116. IEEE (2020)

  8. [7]

    In: IEEE Transactions on Multimedia, vol

    Ye, S., Wang, H., Tan, M., Liu, F.: Recurrent Affine Transformation for Text-to- Image Synthesis. In: IEEE Transactions on Multimedia, vol. 26, pp. 462–473. IEEE (2024)

Show all 31 references
  1. [8]

    In: Proceedings of the 38th International Conference on Machine Learning, Proceedings of Machine Learning Research, vol

    Ramesh, A., Pavlov, M., Goh, G., Gray, S., Voss, C., Radford, A., Chen, M., Sutskever, I.: Zero-Shot Text-to-Image Generation. In: Proceedings of the 38th International Conference on Machine Learning, Proceedings of Machine Learning Research, vol. 139, pp. 8821–8831. PMLR (2021)

  2. [9]

    In: Advances in Neural Infor- mation Processing Systems, vol

    Ding, M., Yang, Z., Hong, W., Zheng, W., Zhou, C., Yin, D., Lin, J., et al.: Cogview: Mastering text-to-image generation via transformers. In: Advances in Neural Infor- mation Processing Systems, vol. 34, pp. 19822–19835. (2021)

  3. [10]

    In: Proceedings of the 31st International Conference on Neural Information Processing Systems (NIPS’17), pp

    van den Oord, A., Vinyals, O., Kavukcuoglu, K.: Neural discrete representation learning. In: Proceedings of the 31st International Conference on Neural Information Processing Systems (NIPS’17), pp. 6309–6318. Curran Associates Inc., Red Hook, NY, USA (2017)

  4. [11]

    arXiv preprint arXiv:2206.10789, vol

    Yu,J.,Xu,Y.,Koh,J.Y.,Luong,T.,Baid,G.,Wang,Z.,Vasudevan,V.,etal.:Scal- ing autoregressive models for content-rich text-to-image generation. arXiv preprint arXiv:2206.10789, vol. 2, no. 3, p. 5 (2022)

  5. [12]

    arXiv preprint arXiv:2110.04627 (2021)

    Yu, J., Li, X., Koh, J.Y., Zhang, H., Pang, R., Qin, J., Ku, A., Xu, Y., Baldridge, J., Wu, Y.: Vector-quantized image modeling with improved VQGAN. arXiv preprint arXiv:2110.04627 (2021)

  6. [13]

    arXiv preprint arXiv:2112.10741 (2021)

    Nichol, A., Dhariwal, P., Ramesh, A., Shyam, P., Mishkin, P., McGrew, B., Sutskever, I., Chen, M.: Glide: Towards photorealistic image generation and editing with text-guided diffusion models. arXiv preprint arXiv:2112.10741 (2021)

  7. [14]

    In: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Gu, S., Liu, Z., Ye, X., Lin, T., Wang, M., Cui, S., Liu, H., Liu, Y., Sun, C., Du, J., Hu, H.: Vector Quantized Diffusion Model for Text-to-Image Synthesis. In: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 10686–10696. IEEE, New Orleans, LA,...

  8. [15]

    arXiv preprint arXiv:2204.06125, vol

    Ramesh, A., Dhariwal, P., Nichol, A., Chu, C., Chen, M.: Hierarchical text- conditional image generation with CLIP latents. arXiv preprint arXiv:2204.06125, vol. 1, no. 2, p. 3 (2022)

  9. [16]

    In: Proceedings of the 36th International Con- ference on Neural Information Processing Systems (NIPS ’22), Article 2643, pp

    Saharia, C., Chan, W., Saxena, S., Lit, L., Whang, J., Denton, E., Seyed Ghasemipour, S.K., Karagol Ayan, B., Mahdavi, S.S., Gontijo-Lopes, R., Salimans, T., Ho, J., Fleet, D.J., Norouzi, M.: Photorealistic text-to-image diffusion models with deep language understanding. In: P...

  10. [17]

    In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Wang, X., Zhang, H., Huang, W., Scott, M.R.: Cross-Batch Memory for Embed- ding Learning. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 6387–6396. IEEE, Seattle, WA, USA (2020)

  11. [19]

    In: Proceedings of the 30th International Conference on Neural Information Processing Systems (NIPS’16), pp

    Salimans, T., Goodfellow, I., Zaremba, W., Cheung, V., Radford, A., Chen, X.: Improved techniques for training GANs. In: Proceedings of the 30th International Conference on Neural Information Processing Systems (NIPS’16), pp. 2234–2242. Curran Associates Inc., Red Hook, NY, USA (2016)

  12. [20]

    In: Proceedings of the 30th International Conference on Neural Information Processing Systems (NIPS’16), pp

    Chen, X., Duan, Y., Houthooft, R., Schulman, J., Sutskever, I., Abbeel, P.: Info- GAN: interpretable representation learning by information maximizing generative adversarial nets. In: Proceedings of the 30th International Conference on Neural Information Processing Systems (NI...

  13. [21]

    In: Proceedings of the 30th International Conference on Neural Information Processing Systems (NIPS’16), pp

    Nguyen, A., Dosovitskiy, A., Yosinski, J., Brox, T., Clune, J.: Synthesizing the preferred inputs for neurons in neural networks via deep generator networks. In: Proceedings of the 30th International Conference on Neural Information Processing Systems (NIPS’16), pp. 3395–3403....

  14. [22]

    In: Proceedings of the 34th International Conference on Machine Learning (ICML’17), vol

    Odena,A.,Olah,C.,Shlens,J.:Conditionalimagesynthesiswithauxiliaryclassifier GANs. In: Proceedings of the 34th International Conference on Machine Learning (ICML’17), vol. 70, pp. 2642–2651. JMLR.org (2017)

  15. [23]

    arXiv preprint arXiv:2107.02423 (2021)

    Ye, H., Yang, X., Takac, M., Sunderraman, R., Ji, S.: Improving text-to-image synthesis using contrastive learning. arXiv preprint arXiv:2107.02423 (2021)

  16. [24]

    arXiv preprint arXiv:1703.06412 (2017)

    Dash, A., Gamboa, J.C.B., Ahmed, S., Liwicki, M., Afzal, M.Z.: Tac-gan-text conditioned auxiliary classifier generative adversarial network. arXiv preprint arXiv:1703.06412 (2017)

  17. [25]

    Ouyang, Y

    X. Ouyang, Y. Chen and G. Agam: Accelerated WGAN update strategy with loss change rate balancing, 2021 IEEE Winter Conference on Applications of Computer Vision (WACV), pp. 2545-2554, Waikoloa, HI, USA. (2020)

  18. [26]

    Ouyang, Y

    X. Ouyang, Y. Chen, K. Zhu and G. Agam: Image restoration refinement with Uformer GAN, 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), pp. 5919-5928, Seattle, WA, USA. (2024)

  19. [27]

    In: 2021 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pp

    Zhang, H., Koh, J.Y., Baldridge, J., Lee, H., Yang, Y.: Cross-Modal Contrastive Learning for Text-to-Image Generation. In: 2021 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pp. 833–842. IEEE, Nashville, TN, USA (2021)

  20. [28]

    In: 2019 IEEE/CVF Conference on Computer Vision 16 X

    Yin, G., Liu, B., Sheng, L., Yu, N., Wang, X., Shao, J.: Semantics Disentangling for Text-To-Image Generation. In: 2019 IEEE/CVF Conference on Computer Vision 16 X. Ouyang et al. and Pattern Recognition (CVPR), pp. 2322–2331. IEEE, Long Beach, CA, USA (2019)

  21. [29]

    In: Advances in Neural Information Pro- cessing Systems 29: Annual Conference on Neural Information Processing Systems 2016, pp

    Salimans, T., Goodfellow, I.J., Zaremba, W., Cheung, V., Radford, A., Chen, X.: Improved Techniques for Training GANs. In: Advances in Neural Information Pro- cessing Systems 29: Annual Conference on Neural Information Processing Systems 2016, pp. 2226–2234. (2016)

  22. [30]

    Journal of Multivariate Analysis, vol

    Dowson, D.C., Landau, B.V.: The Fréchet distance between multivariate normal distributions. Journal of Multivariate Analysis, vol. 12, no. 3, pp. 450–455. (1982)

  23. [31]

    arXiv preprint arXiv:1801.01973

    Barratt, S., Sharma, R.: A Note on the Inception Score. arXiv preprint arXiv:1801.01973. (2018). 6 Appendix 6.1 Comparision results We compare with the DALLE-2 and Stable Diffusion which are the most popu- lar models for text to image synthesis task. Since neither DALLE-2 nor ...

Pith tools

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