pith. sign in

arxiv: 2606.05068 · v1 · pith:JNQQVM54new · submitted 2026-06-03 · 💻 cs.CV

MaCo-GAN: Manifold-Contrastive Adversarial Learning for Single Image Super-Resolution

Daeyoung Han , Seongmin Hwang , Moongu Jeon This is my paper

Pith reviewed 2026-06-28 06:51 UTC · model grok-4.3

classification 💻 cs.CV
keywords single image super-resolutionGANcontrastive learningmanifold contrastiveperception-distortion tradeoffadversarial trainingfake sample synthesis
0
0 comments X

The pith

Replacing the adversarial loss in super-resolution GANs with a manifold-contrastive objective improves the perception-distortion trade-off.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that conventional GAN discriminators for single-image super-resolution focus on overall naturalness instead of conditional realism, leading to artifacts. It introduces a dynamic fake sample synthesizer that converts ground-truth images into a range of perceptually plausible fakes while preserving exact low-resolution correspondence. These samples enable a supervised contrastive minimax game in which the generator pulls its outputs toward on-manifold fakes and away from off-manifold ones, with the discriminator doing the reverse. Substituting this objective for a baseline model's standard adversarial loss produces consistent gains on the perception-distortion frontier across benchmarks.

Core claim

By replacing the conventional adversarial loss with a supervised contrastive objective built on a dynamic fake sample synthesizer, the generator is trained to attract predictions to on-manifold fakes and repel them from off-manifold fakes while the discriminator optimizes the opposite, yielding measurable improvements in the perception-distortion trade-off when inserted into existing single-image super-resolution models.

What carries the argument

The dynamic fake sample synthesizer, which generates a spectrum of challenging yet low-resolution-corresponding fake images from ground truth to support the conditional contrastive minimax game.

If this is right

  • Any existing GAN-based super-resolution model can adopt the objective by loss substitution alone and expect improved conditional realism.
  • The contrastive game enforces stricter low-resolution fidelity than standard adversarial training because synthesized fakes are explicitly LR-matched.
  • Discriminator training becomes a direct optimization over manifold distance rather than a binary real/fake decision.
  • The framework supplies a controllable spectrum of positive and negative examples that scale with the desired distortion level.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same synthesizer-plus-contrastive pattern could be tested in other conditional image-to-image tasks such as denoising or deblurring where correspondence to input must be preserved.
  • If the on-manifold and off-manifold distinction proves robust, it may reduce the need for perceptual loss terms that currently require separate pretrained networks.
  • Extending the synthesizer to produce multi-scale or patch-level fakes might further tighten the conditional manifold constraint.

Load-bearing premise

The dynamic fake sample synthesizer can produce perceptually plausible fake images that maintain strict low-resolution correspondence while spanning a useful range of distortion levels.

What would settle it

Running the same baseline super-resolution models with and without the contrastive objective on standard benchmarks and checking whether the perception-distortion Pareto front shifts measurably when the synthesizer is ablated or replaced by random perturbations.

Figures

Figures reproduced from arXiv: 2606.05068 by Daeyoung Han, Moongu Jeon, Seongmin Hwang.

Figure 1
Figure 1. Figure 1: Graphical comparison of our proposed contrastive GAN with existing frame￾works in SISR. by forcing the model to find the averages of plausible solutions, thereby collaps￾ing perceptual variance [18, 25, 26]. To overcome this limitation, Generative Adversarial Networks (GANs) [10] were introduced to the field of perceptual SISR. [24] By bringing a generator (an SISR network) and a discriminator into a minim… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of our proposed framework. (a) The fake sample synthesizer ψ(·, σ) yields random fake samples from the noisy GT, which is corrupted by additive random Gaussian noise with a level of σ. The outputs exhibit diverse high-frequency details while conserving the LR correspondence with the input GT. (b) To establish a con￾trastive learning process, the sets of proper keys for the query (SR prediction) mu… view at source ↗
Figure 3
Figure 3. Figure 3: Visualization of fake samples generated by our proposed model with varying levels of distortion, controlled by σ. As σ increases from 0.0 (the GT) to 5.0, the samples become progressively more distorted. Notably, even at the most severe deformation, a certain level of LR-PSNR and LPIPS scores is maintained. This demonstrates that the fake samples satisfy both photo-realism and LR correspondence, confirming… view at source ↗
Figure 4
Figure 4. Figure 4: Visual comparison between our proposed method and previous models with the RRDB backbone [46] for ×4 upscaling. Zoom in for the best view [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 1
Figure 1. Figure 1: Quantitative analysis of our syn￾thesized fakes. The applied noise level (σ) shows a significant correlation with the perceptual distance from the GT (LPIPS, red), validating its effectiveness in modulating the difficulty of the out￾puts while maintaining high LR corre￾spondence (LR-PSNR, blue). 10 8 6 4 2 0 2 4 2 0 2 4 [PITH_FULL_IMAGE:figures/full_fig_p020_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Visualization of our proposed model and the baseline methods for the bicubic ×4 SR task. Zoom in for the best view [PITH_FULL_IMAGE:figures/full_fig_p023_3.png] view at source ↗
read the original abstract

Conventional Generative Adversarial Networks (GANs) for Single Image Super-Resolution (SISR) often struggle with hallucinated artifacts, largely because standard discriminators evaluate overall image naturalness rather than strict conditional realism. To address this, we propose MaCo-GAN, a novel manifold-contrastive GAN framework that replaces the conventional adversarial loss with a supervised contrastive objective. A core component of our method is a dynamic fake sample synthesizer that transforms ground truth (GT) data into a spectrum of challenging, perceptually plausible fake images that strictly maintain low-resolution (LR) correspondence. Utilizing these synthesized samples, we establish a robust contrastive minimax game: the generator is trained to attract its predictions toward on-manifold fakes (low distortion) and repel them from off-manifold fakes (high distortion), while the discriminator optimizes the exact opposite. By simply replacing the adversarial loss of a baseline SR model with our proposed objective, we demonstrate consistent improvements in the perception-distortion trade-off across various benchmarks. Extensive ablation studies validate the effectiveness of our framework and provide deep insights into the dynamics of this conditional contrastive game.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript proposes MaCo-GAN, a manifold-contrastive GAN for single-image super-resolution. It replaces the standard adversarial loss with a supervised contrastive objective built around a dynamic fake-sample synthesizer that converts ground-truth images into a spectrum of on-manifold (low-distortion) and off-manifold (high-distortion) fakes while preserving strict low-resolution correspondence. The generator is trained to attract its outputs toward on-manifold fakes and repel them from off-manifold fakes; the discriminator does the reverse. The central claim is that simply substituting this objective for the adversarial loss in any baseline SR model yields consistent gains on the perception-distortion trade-off across benchmarks, with supporting ablation studies.

Significance. If the empirical claims are substantiated, the framework would supply a concrete mechanism for enforcing conditional realism inside conditional GANs for SISR by explicitly contrasting synthesized manifold samples. The dynamic synthesizer is presented as the key technical device enabling the contrastive minimax game. The approach directly targets a recognized weakness of conventional SR discriminators and, if the reported gains prove robust, would constitute a useful incremental advance in the perception-distortion literature.

major comments (2)
  1. [Abstract] Abstract, final sentence: the assertion that "simply replacing the adversarial loss of a baseline SR model with our proposed objective" produces "consistent improvements in the perception-distortion trade-off across various benchmarks" is unsupported by any quantitative metrics, error bars, tables, figures, dataset descriptions, or ablation results in the supplied text, rendering the central empirical claim unevaluable.
  2. [Abstract] Abstract, paragraph 2: the dynamic fake-sample synthesizer is described as transforming GT data into "challenging, perceptually plausible fake images that strictly maintain low-resolution correspondence," yet no equations, algorithmic steps, or independence arguments are supplied to show that the synthesis procedure itself is free of fitted parameters or circular dependence on the model being trained; this directly affects whether the contrastive objective can be evaluated as a drop-in replacement.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive feedback on the abstract. We address each major comment below and indicate the revisions we will make to improve clarity and evaluability while preserving the manuscript's core contributions.

read point-by-point responses

  1. Referee: [Abstract] Abstract, final sentence: the assertion that "simply replacing the adversarial loss of a baseline SR model with our proposed objective" produces "consistent improvements in the perception-distortion trade-off across various benchmarks" is unsupported by any quantitative metrics, error bars, tables, figures, dataset descriptions, or ablation results in the supplied text, rendering the central empirical claim unevaluable.

    Authors: We acknowledge that the abstract, being a concise summary, does not embed the specific numerical results, error bars, or table references. The full manuscript contains the supporting quantitative evidence, including tables, figures, dataset details, and ablation studies across benchmarks that substantiate the claim of consistent improvements. To directly address the evaluability concern, we will revise the abstract to incorporate key quantitative metrics (e.g., average gains in perceptual and distortion measures) or add explicit cross-references to the relevant results sections and tables. This change will make the empirical claim traceable from the abstract itself. revision: yes

  2. Referee: [Abstract] Abstract, paragraph 2: the dynamic fake-sample synthesizer is described as transforming GT data into "challenging, perceptually plausible fake images that strictly maintain low-resolution correspondence," yet no equations, algorithmic steps, or independence arguments are supplied to show that the synthesis procedure itself is free of fitted parameters or circular dependence on the model being trained; this directly affects whether the contrastive objective can be evaluated as a drop-in replacement.

    Authors: The abstract provides only a high-level description of the synthesizer. The full manuscript supplies the complete equations, algorithmic steps, and arguments demonstrating that the procedure is parameter-free and maintains strict independence from the trained SR model (no circular dependence), thereby supporting its use as a drop-in replacement. To improve evaluability at the abstract level, we will add a concise clarification that the synthesizer operates without learned parameters and preserves LR correspondence independently, with full technical details provided in the methods section. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The abstract and supplied description introduce a contrastive objective built on a dynamic fake-sample synthesizer, but present no equations, derivations, or self-citations that reduce any claimed result to its own inputs by construction. The central claim is an empirical observation that replacing the adversarial loss yields perception-distortion gains; this is a testable performance statement rather than a self-referential prediction or fitted-input renaming. No load-bearing step matches any of the enumerated circularity patterns, and the method is described as self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Review performed on abstract only; no explicit free parameters, axioms, or invented entities are stated in the provided text.

pith-pipeline@v0.9.1-grok · 5737 in / 1032 out tokens · 38794 ms · 2026-06-28T06:51:04.014735+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

56 extracted references · 14 canonical work pages · 8 internal anchors

  1. [1]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition workshops

    Agustsson, E., Timofte, R.: Ntire 2017 challenge on single image super-resolution: Dataset and study. In: Proceedings of the IEEE conference on computer vision and pattern recognition workshops. pp. 126–135 (2017)

    2017

  2. [2]

    In: European Conference on Computer Vision

    Ahn, D., Cho, H., Min, J., Jang, W., Kim, J., Kim, S., Park, H.H., Jin, K.H., Kim, S.: Self-rectifying diffusion sampling with perturbed-attention guidance. In: European Conference on Computer Vision. pp. 1–17. Springer (2024)

    2024

  3. [3]

    TripletGAN: Training Generative Model with Triplet Loss

    Cao,G.,Yang,Y.,Lei,J.,Jin,C.,Liu,Y.,Song,M.:Tripletgan:Traininggenerative model with triplet loss. arXiv preprint arXiv:1711.05084 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  4. [4]

    In: International conference on machine learning

    Chen, T., Kornblith, S., Norouzi, M., Hinton, G.: A simple framework for con- trastive learning of visual representations. In: International conference on machine learning. pp. 1597–1607. PmLR (2020)

    2020

  5. [5]

    Improved Baselines with Momentum Contrastive Learning

    Chen, X., Fan, H., Girshick, R., He, K.: Improved baselines with momentum con- trastive learning. arXiv preprint arXiv:2003.04297 (2020)

    work page internal anchor Pith review Pith/arXiv arXiv 2003

  6. [6]

    Advances in neural information processing systems34, 8780–8794 (2021)

    Dhariwal, P., Nichol, A.: Diffusion models beat gans on image synthesis. Advances in neural information processing systems34, 8780–8794 (2021)

    2021

  7. [7]

    IEEE transactions on pattern analysis and machine intelligence44(5), 2567–2581 (2020)

    Ding, K., Ma, K., Wang, S., Simoncelli, E.P.: Image quality assessment: Unify- ing structure and texture similarity. IEEE transactions on pattern analysis and machine intelligence44(5), 2567–2581 (2020)

    2020

  8. [8]

    In: European conference on computer vision

    Dong, C., Loy, C.C., Tang, X.: Accelerating the super-resolution convolutional neural network. In: European conference on computer vision. pp. 391–407. Springer (2016)

    2016

  9. [9]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Esser, P., Rombach, R., Ommer, B.: Taming transformers for high-resolution image synthesis. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 12873–12883 (2021)

    2021

  10. [10]

    Advances in neural in- formation processing systems27(2014)

    Goodfellow, I.J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., Bengio, Y.: Generative adversarial nets. Advances in neural in- formation processing systems27(2014)

    2014

  11. [11]

    In: Proceedings of the thirteenth inter- national conference on artificial intelligence and statistics

    Gutmann, M., Hyvärinen, A.: Noise-contrastive estimation: A new estimation prin- ciple for unnormalized statistical models. In: Proceedings of the thirteenth inter- national conference on artificial intelligence and statistics. pp. 297–304. JMLR Workshop and Conference Proceedings (2010)

    2010

  12. [12]

    Momentum contrast for unsupervised visual representation learning,

    He, K., Fan, H., Wu, Y., Xie, S., Girshick, R.: Momentum contrast for unsupervised visual representation learning. arxiv e-prints, art. arXiv preprint arXiv:1911.05722 2(2019)

    work page arXiv 1911

  13. [13]

    Advances in neural information processing systems33, 6840–6851 (2020)

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. Advances in neural information processing systems33, 6840–6851 (2020)

    2020

  14. [14]

    Classifier-Free Diffusion Guidance

    Ho, J., Salimans, T.: Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598 (2022)

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  15. [15]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Hong, S., Lee, G., Jang, W., Kim, S.: Improving sample quality of diffusion mod- els using self-attention guidance. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 7462–7471 (2023) 16 D. Han, et al

    2023

  16. [16]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Huang, J.B., Singh, A., Ahuja, N.: Single image super-resolution from transformed self-exemplars. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 5197–5206 (2015)

    2015

  17. [17]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Isola, P., Zhu, J.Y., Zhou, T., Efros, A.A.: Image-to-image translation with condi- tional adversarial networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 1125–1134 (2017)

    2017

  18. [18]

    Machine learning51(2), 115–135 (2003)

    James, G.M.: Variance and bias for general loss functions. Machine learning51(2), 115–135 (2003)

    2003

  19. [19]

    Training gans with stronger augmen- tations via contrastive discriminator,

    Jeong, J., Shin, J.: Training gans with stronger augmentations via contrastive discriminator. arXiv preprint arXiv:2103.09742 (2021)

    work page arXiv 2021

  20. [20]

    The relativistic discriminator: a key element missing from standard GAN

    Jolicoeur-Martineau,A.:Therelativisticdiscriminator:Akeyelementmissingfrom standard gan. arXiv preprint arXiv:1807.00734 (2018)

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  21. [21]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Ke, J., Wang, Q., Wang, Y., Milanfar, P., Yang, F.: Musiq: Multi-scale image quality transformer. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 5148–5157 (2021)

    2021

  22. [22]

    Advances in neural information processing systems33, 18661–18673 (2020)

    Khosla, P., Teterwak, P., Wang, C., Sarna, A., Tian, Y., Isola, P., Maschinot, A., Liu, C., Krishnan, D.: Supervised contrastive learning. Advances in neural information processing systems33, 18661–18673 (2020)

    2020

  23. [23]

    arXiv preprint arXiv:2501.17683 (2025)

    Kim, B.J., Kim, S.W.: Temperature-free loss function for contrastive learning. arXiv preprint arXiv:2501.17683 (2025)

    work page arXiv 2025

  24. [24]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Ledig, C., Theis, L., Huszár, F., Caballero, J., Cunningham, A., Acosta, A., Aitken, A., Tejani, A., Totz, J., Wang, Z., et al.: Photo-realistic single image super- resolution using a generative adversarial network. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 4681–4690 (2017)

    2017

  25. [25]

    Lee, M., Hyun, S., Jun, W., Heo, J.P.: Auto-encoded supervision for perceptual imagesuper-resolution.In:ProceedingsoftheComputerVisionandPatternRecog- nition Conference. pp. 17958–17968 (2025)

    2025

  26. [26]

    Harmonizing Maximum Likelihood with GANs for Multimodal Conditional Generation

    Lee, S., Ha, J., Kim, G.: Harmonizing maximum likelihood with gans for multi- modal conditional generation. arXiv preprint arXiv:1902.09225 (2019)

    work page internal anchor Pith review Pith/arXiv arXiv 1902

  27. [27]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Li, Y., Zhang, K., Liang, J., Cao, J., Liu, C., Gong, R., Zhang, Y., Tang, H., Liu, Y., Demandolx, D., et al.: Lsdir: A large scale dataset for image restoration. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 1775–1787 (2023)

    2023

  28. [28]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Liang, J., Zeng, H., Zhang, L.: Details or artifacts: A locally discriminative learning approach to realistic image super-resolution. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 5657–5666 (2022)

    2022

  29. [29]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition workshops

    Lim,B.,Son,S.,Kim,H.,Nah,S.,MuLee,K.:Enhanceddeepresidualnetworksfor single image super-resolution. In: Proceedings of the IEEE conference on computer vision and pattern recognition workshops. pp. 136–144 (2017)

    2017

  30. [30]

    Geometric GAN

    Lim, J.H., Ye, J.C.: Geometric gan. arXiv preprint arXiv:1705.02894 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  31. [31]

    IEEE transactions on pattern analysis and machine intelligence44(11), 7898–7911 (2021)

    Ma, C., Rao, Y., Lu, J., Zhou, J.: Structure-preserving image super-resolution. IEEE transactions on pattern analysis and machine intelligence44(11), 7898–7911 (2021)

    2021

  32. [32]

    In: Proceedings eighth IEEE international conference on com- puter vision

    Martin, D., Fowlkes, C., Tal, D., Malik, J.: A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics. In: Proceedings eighth IEEE international conference on com- puter vision. ICCV 2001. vol. 2, pp. 416–423. IEEE (2001)

    2001

  33. [33]

    Multimedia tools and applications76(20), 21811–21838 (2017) MaCo-GAN 17

    Matsui, Y., Ito, K., Aramaki, Y., Fujimoto, A., Ogawa, T., Yamasaki, T., Aizawa, K.: Sketch-based manga retrieval using manga109 dataset. Multimedia tools and applications76(20), 21811–21838 (2017) MaCo-GAN 17

    2017

  34. [34]

    In: Proceedings of the ieee/cvf conference on computer vision and pattern recognition

    Menon, S., Damian, A., Hu, S., Ravi, N., Rudin, C.: Pulse: Self-supervised photo upsampling via latent space exploration of generative models. In: Proceedings of the ieee/cvf conference on computer vision and pattern recognition. pp. 2437–2445 (2020)

    2020

  35. [35]

    completely blind

    Mittal, A., Soundararajan, R., Bovik, A.C.: Making a “completely blind” image quality analyzer. IEEE Signal processing letters20(3), 209–212 (2012)

    2012

  36. [36]

    Advances in Neural Information Processing Systems34, 16398–16409 (2021)

    Ning, Q., Dong, W., Li, X., Wu, J., Shi, G.: Uncertainty-driven loss for single image super-resolution. Advances in Neural Information Processing Systems34, 16398–16409 (2021)

    2021

  37. [37]

    In: International conference on machine learning

    Odena, A., Olah, C., Shlens, J.: Conditional image synthesis with auxiliary classi- fier gans. In: International conference on machine learning. pp. 2642–2651. PMLR (2017)

    2017

  38. [38]

    Representation Learning with Contrastive Predictive Coding

    Oord, A.v.d., Li, Y., Vinyals, O.: Representation learning with contrastive predic- tive coding. arXiv preprint arXiv:1807.03748 (2018)

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  39. [39]

    Advances in neural information processing sys- tems32(2019)

    Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., et al.: Pytorch: An imperative style, high- performance deep learning library. Advances in neural information processing sys- tems32(2019)

    2019

  40. [40]

    In: International Conference on Medical image computing and computer-assisted intervention

    Ronneberger, O., Fischer, P., Brox, T.: U-net: Convolutional networks for biomedi- cal image segmentation. In: International Conference on Medical image computing and computer-assisted intervention. pp. 234–241. Springer (2015)

    2015

  41. [41]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Schonfeld, E., Schiele, B., Khoreva, A.: A u-net based discriminator for generative adversarial networks. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 8207–8216 (2020)

    2020

  42. [42]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Soh, J.W., Park, G.Y., Jo, J., Cho, N.I.: Natural and realistic single image super- resolution with explicit natural manifold discrimination. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 8122–8131 (2019)

    2019

  43. [43]

    In: Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition

    Tseng, H.Y., Jiang, L., Liu, C., Yang, M.H., Yang, W.: Regularizing generative adversarial networks under limited data. In: Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition. pp. 7921–7931 (2021)

    2021

  44. [44]

    In: Proceedings of the AAAI conference on artificial intelligence

    Wang, J., Chan, K.C., Loy, C.C.: Exploring clip for assessing the look and feel of images. In: Proceedings of the AAAI conference on artificial intelligence. vol. 37, pp. 2555–2563 (2023)

    2023

  45. [45]

    In: Proceedings of the IEEE/CVF in- ternational conference on computer vision

    Wang, X., Xie, L., Dong, C., Shan, Y.: Real-esrgan: Training real-world blind super-resolution with pure synthetic data. In: Proceedings of the IEEE/CVF in- ternational conference on computer vision. pp. 1905–1914 (2021)

    1905

  46. [46]

    In: Proceed- ings of the European conference on computer vision (ECCV) workshops

    Wang, X., Yu, K., Wu, S., Gu, J., Liu, Y., Dong, C., Qiao, Y., Change Loy, C.: Esrgan: Enhanced super-resolution generative adversarial networks. In: Proceed- ings of the European conference on computer vision (ECCV) workshops. pp. 0–0 (2018)

    2018

  47. [47]

    IEEE transactions on pattern analysis and machine intelligence43(10), 3365–3387 (2020)

    Wang, Z., Chen, J., Hoi, S.C.: Deep learning for image super-resolution: A survey. IEEE transactions on pattern analysis and machine intelligence43(10), 3365–3387 (2020)

    2020

  48. [48]

    arXiv preprint arXiv:2409.16211 , year=

    Weber, M., Yu, L., Yu, Q., Deng, X., Shen, X., Cremers, D., Chen, L.C.: Maskbit: Embedding-free image generation via bit tokens. arXiv preprint arXiv:2409.16211 (2024)

    work page arXiv 2024

  49. [49]

    IEEE Transactions on Neural Networks and Learning Sys- tems35(11), 15834–15845 (2023) 18 D

    Wu, G., Jiang, J., Liu, X.: A practical contrastive learning framework for single- image super-resolution. IEEE Transactions on Neural Networks and Learning Sys- tems35(11), 15834–15845 (2023) 18 D. Han, et al

    2023

  50. [50]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Wu, Z., Xiong, Y., Yu, S.X., Lin, D.: Unsupervised feature learning via non- parametric instance discrimination. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 3733–3742 (2018)

    2018

  51. [51]

    IEEE Transactions on Pattern Analysis and Machine Intelligence46(9), 6199–6215 (2024)

    Xu,T.,Li,L.,Mi,P.,Zheng,X.,Chao,F.,Ji,R.,Tian,Y.,Shen,Q.:Uncoveringthe over-smoothing challenge in image super-resolution: Entropy-based quantification and contrastive optimization. IEEE Transactions on Pattern Analysis and Machine Intelligence46(9), 6199–6215 (2024)

    2024

  52. [52]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recog- nition

    Yang, S., Wu, T., Shi, S., Lao, S., Gong, Y., Cao, M., Wang, J., Yang, Y.: Maniqa: Multi-dimension attention network for no-reference image quality assessment. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recog- nition. pp. 1191–1200 (2022)

    2022

  53. [53]

    Language Model Beats Diffusion -- Tokenizer is Key to Visual Generation

    Yu, L., Lezama, J., Gundavarapu, N.B., Versari, L., Sohn, K., Minnen, D., Cheng, Y., Birodkar, V., Gupta, A., Gu, X., et al.: Language model beats diffusion– tokenizer is key to visual generation. arXiv preprint arXiv:2310.05737 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  54. [54]

    Advances in Neural Information Processing Systems37, 128940–128966 (2024)

    Yu, Q., Weber, M., Deng, X., Shen, X., Cremers, D., Chen, L.C.: An image is worth 32 tokens for reconstruction and generation. Advances in Neural Information Processing Systems37, 128940–128966 (2024)

    2024

  55. [55]

    arXiv preprint arXiv:2107.00708 (2021)

    Zhang, J., Lu, S., Zhan, F., Yu, Y.: Blind image super-resolution via contrastive representation learning. arXiv preprint arXiv:2107.00708 (2021)

    work page arXiv 2021

  56. [56]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Zhang, R., Isola, P., Efros, A.A., Shechtman, E., Wang, O.: The unreasonable effectiveness of deep features as a perceptual metric. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 586–595 (2018) Supplementary Material for MaCo-GAN: Manifold-Contrastive Adversarial Learning for Single Image Super-Resolution Daeyoung H...

    work page arXiv 2018