Pith. sign in

REVIEW 2 major objections 1 minor 41 references

Training a small CNN as a fixed-point operator lets repeated passes improve blind face restoration without extra inference cost.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-06-27 22:50 UTC pith:MEGEDD27

load-bearing objection CFRNet shows a deployable small-network trick for multi-pass face restoration on NPUs, but the multi-cycle gains rest on unverified test-set construction. the 2 major comments →

arxiv 2606.06850 v1 pith:MEGEDD27 submitted 2026-06-05 cs.CV

CFRNet: Cycle-Consistent Fixed-Point Training for Real-Time Blind Face Restoration on Consumer Embedded NPUs

Fuchen Li , Xinyang Wang , Yahui Zhang , Yuhan Chen , Jiahong Guo , Zhuohan Qin , Wenbo Ma This is my paper
classification cs.CV
keywords blind face restorationcycle-consistent trainingfixed-point operatoron-device NPUembedded inferencereal-time image restorationlightweight CNN
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper introduces CFRNet, a 2-million-parameter ResNet-style network designed for 256x256 face restoration on small consumer NPUs. Instead of one-pass training followed by manual repetition, it uses Cycle-Consistent Fixed-Point Training with progressive multi-cycle supervision, an idempotence loss, and a re-degradation cycle loss so the network learns to leave its own outputs unchanged. This property turns the cycle count into a runtime quality knob that needs no retraining. On a 300-image test set the method records the lowest LPIPS at three cycles and best PSNR and SSIM at two cycles while running at 23 ms per cycle in INT8 on a HiSilicon Hi3402 NPU; the same baselines cannot be compiled to that hardware. The same training recipe also succeeds with a plain CNN and supports real-time use on an in-car driver-monitoring board.

Core claim

CFRNet trains the network to function as a fixed-point operator so that repeated application on a restored face produces no further change. The three losses enforce consistency across cycles and through re-degradation, allowing the model to refine outputs at inference time by simply iterating the same weights.

What carries the argument

Cycle-Consistent Fixed-Point Training (CCFP), which combines multi-cycle supervision, idempotence, and re-degradation losses to make the network converge to stable outputs under repeated application.

Load-bearing premise

Repeated application of the trained network improves quality on real degraded faces outside the training distribution rather than merely satisfying the training losses.

What would settle it

Measure whether LPIPS, PSNR, and visual artifacts continue to improve or degrade when the model is applied four or more times to a fresh collection of real-world degraded face photographs never seen during training.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • PSNR peaks at two cycles and LPIPS keeps improving through three cycles on the reported test set.
  • The number of cycles acts as a simple quality control parameter that requires no retraining or architecture change.
  • The same fixed-point training recipe produces usable results with a plain CNN that is even simpler to deploy.
  • The model achieves real-time INT8 inference on the HiSilicon Hi3402 NPU and on an in-car driver-monitoring board.

Where Pith is reading between the lines

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

  • The approach may reduce reliance on large generative priors for on-device restoration tasks if the fixed-point property transfers to other image domains.
  • A direct test on non-face restoration problems such as super-resolution or denoising would show whether the training recipe is specific to faces or more general.
  • Measuring memory and latency on additional embedded NPUs would clarify how widely the compile-time advantage holds.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The manuscript presents CFRNet, a lightweight 2M-parameter ResNet-style network for blind face restoration at 256x256 resolution on consumer NPUs. It introduces Cycle-Consistent Fixed-Point Training (CCFP) with three losses—progressive multi-cycle supervision, idempotence loss, and re-degradation cycle loss—to train the network as a fixed-point operator. This allows multiple inference cycles to improve quality (e.g., LPIPS dropping 31% from 1 to 3 cycles) without additional training. The method is compared to retrained baselines on a 300-image test set, showing superior metrics, and demonstrates real-time INT8 performance on HiSilicon Hi3402 NPU, with cycle count serving as a quality knob.

Significance. If the reported gains hold and generalize, the work offers a practical solution for deploying face restoration on resource-constrained embedded devices where generative priors are infeasible. The fixed-point training enabling tunable quality via repeated application is a notable contribution for real-time applications like in-car monitoring.

major comments (2)
  1. Abstract: The 300-image test set is not described in terms of its construction, the degradation pipeline used, or its relation to the training data. This information is essential to evaluate whether the multi-cycle improvements (LPIPS 0.250 at three cycles vs. one cycle) demonstrate generalization of the fixed-point property to real blind degradations or are limited to in-distribution synthetic cases, which is central to the paper's claims.
  2. Abstract: Details on the training data, exact implementations of the retrained baselines, and any statistical significance testing for the metric improvements are missing, weakening the support for the performance claims.
minor comments (1)
  1. Abstract: Notation inconsistencies such as '2.0,M-parameter' and '23,ms' should be corrected to '2.0M-parameter' and '23 ms'.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments highlighting the need for greater transparency in experimental details. These points are valid and we will revise the manuscript to incorporate the requested information, strengthening the presentation of our results without altering the core claims.

read point-by-point responses

  1. Referee: Abstract: The 300-image test set is not described in terms of its construction, the degradation pipeline used, or its relation to the training data. This information is essential to evaluate whether the multi-cycle improvements (LPIPS 0.250 at three cycles vs. one cycle) demonstrate generalization of the fixed-point property to real blind degradations or are limited to in-distribution synthetic cases, which is central to the paper's claims.

    Authors: We agree that a detailed description of the test set is essential to substantiate the generalization claims. In the revised manuscript we will add a dedicated paragraph in the Experiments section describing the 300-image test set construction (image sources, selection criteria, and resolution), the exact degradation pipeline (including noise, blur, and compression parameters), and its deliberate separation from the training distribution to confirm that multi-cycle gains reflect the fixed-point property on unseen blind degradations rather than in-distribution overfitting. revision: yes

  2. Referee: Abstract: Details on the training data, exact implementations of the retrained baselines, and any statistical significance testing for the metric improvements are missing, weakening the support for the performance claims.

    Authors: We acknowledge these omissions limit reproducibility and claim strength. The revision will expand the training data description (datasets, sizes, and augmentation), provide precise baseline implementation details (architectures, loss functions, training schedules, and how each was retrained from scratch at 256x256), and add statistical significance tests (e.g., paired t-tests across the 300 images) for the reported LPIPS, PSNR, and SSIM improvements to quantify that the gains are statistically reliable. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper defines CCFP training via three explicit losses (progressive multi-cycle supervision, idempotence loss, re-degradation cycle loss) and reports empirical metrics on a held-out 300-image test set. No derivation step reduces the claimed multi-cycle gains or fixed-point property to a fitted input, self-definition, or self-citation chain by construction. The method is presented as an independent empirical technique with separate evaluation, satisfying the default expectation of no circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The method rests on standard supervised learning assumptions plus the novel training objectives; no free parameters beyond ordinary network weights are identified, and no new physical entities are introduced.

axioms (1)
  • domain assumption A convolutional network can be trained to approximate a fixed-point operator for the blind restoration mapping.
    This is the core premise that allows the idempotence and cycle losses to produce useful iterative behavior.

pith-pipeline@v0.9.1-grok · 5951 in / 1294 out tokens · 34841 ms · 2026-06-27T22:50:26.135841+00:00 · methodology

0 comments
read the original abstract

Blind face restoration on consumer devices has to balance image quality against speed and memory. Strong methods such as GFPGAN and CodeFormer give good perceptual quality, but they rely on large pretrained generative priors and on operators such as attention, codebook lookup, and style modulation that are hard to compile and quantize on the small neural processing units (NPUs) used in consumer hardware. Small convolutional restorers run fast enough, but they tend to over-smooth and to leave artifacts around the eyes, nose, and mouth. We present CFRNet, a 2.0,M-parameter ResNet-style restorer for on-device use at $256\times256$, the common face-crop size on consumer NPUs. The main idea is Cycle-Consistent Fixed-Point Training (CCFP). Instead of training the network for one pass and then running it several times by hand, we train it to act as a fixed-point operator, so that applying it again to a restored face does not change the face. CCFP uses three training losses, namely progressive multi-cycle supervision, an idempotence loss, and a re-degradation cycle loss, and it adds no cost at inference. To compare fairly under our deployment limits, we retrain all baselines from scratch at the same $256\times256$ resolution. On a 300-image test set, CFRNet reaches the best perceptual score (LPIPS 0.250 at three cycles, which is 31% lower than one cycle) and also the best PSNR and SSIM at two cycles. It runs in about 23,ms per cycle in INT8 on a HiSilicon Hi3402 NPU, while the same baselines cannot be compiled to that chip. The cycle count $k$ acts as a simple quality knob that needs no retraining: PSNR is best at $k\!=\!2$ and LPIPS keeps improving up to $k\!=\!3$. We further show that the same idea works with a plain CNN that is even easier to deploy, and we run the model in real time on an in-car driver-monitoring board.

Figures

Figures reproduced from arXiv: 2606.06850 by Fuchen Li, Jiahong Guo, Wenbo Ma, Xinyang Wang, Yahui Zhang, Yuhan Chen, Zhuohan Qin.

Figure 1
Figure 1. Figure 1: CFRNet and CCFP training. The same 2.0 M-parameter generator () [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Distance to the clean image as a function of the cycle count. A single [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison on the FFHQ-256 test set. Left to right: degraded input, CodeFormer-Lite, GPEN-Lite, GFPGAN-Lite, CFRNet ( [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Cycle-by-cycle output of one trained CFRNet. Left to right: degraded input, [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Component-level comparison with zoom-ins. Yellow box: eye region; [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The plain variant on a hard real-world input (heavy haze, low contrast, [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗

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

41 extracted references

  1. [1]

    Real-time driver drowsiness detection for embedded system using model compression of deep neural networks,

    B. Reddy, Y .-H. Kim, S. Yun, C. Seo, and J. Jang, “Real-time driver drowsiness detection for embedded system using model compression of deep neural networks,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2017, pp. 438–445

    2017

  2. [2]

    Driver drowsiness detection using condition-adaptive representation learning framework,

    J. Yu, S. Park, S. Lee, and M. Jeon, “Driver drowsiness detection using condition-adaptive representation learning framework,”IEEE Transac- tions on Intelligent Transportation Systems, vol. 20, no. 11, pp. 4206– 4218, 2019

    2019

  3. [3]

    Towards real-world blind face restoration with generative facial prior,

    X. Wang, Y . Li, H. Zhang, and Y . Shan, “Towards real-world blind face restoration with generative facial prior,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021, pp. 9168–9178

    2021

  4. [4]

    Gan prior embedded network for blind face restoration in the wild,

    T. Yang, P. Ren, X. Xie, and L. Zhang, “Gan prior embedded network for blind face restoration in the wild,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021, pp. 672–681

    2021

  5. [5]

    Restoreformer: High-quality blind face restoration from undegraded key-value pairs,

    Z. Wang, J. Zhang, R. Chen, W. Wang, and P. Luo, “Restoreformer: High-quality blind face restoration from undegraded key-value pairs,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 17 512–17 521

    2022

  6. [6]

    Towards robust blind face restoration with codebook lookup transformer,

    S. Zhou, K. C. Chan, C. Li, and C. C. Loy, “Towards robust blind face restoration with codebook lookup transformer,” inAdvances in Neural Information Processing Systems (NeurIPS), vol. 35, 2022, pp. 30 599– 30 611

    2022

  7. [7]

    Mobilenetv2: Inverted residuals and linear bottlenecks,

    M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, “Mobilenetv2: Inverted residuals and linear bottlenecks,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 4510–4520

    2018

  8. [8]

    Residual feature distillation network for lightweight image super-resolution,

    J. Liu, J. Tang, and G. Wu, “Residual feature distillation network for lightweight image super-resolution,” inEuropean Conference on Computer Vision (ECCV) Workshops, 2020, pp. 41–55

    2020

  9. [9]

    Deep back-projection networks for super-resolution,

    M. Haris, G. Shakhnarovich, and N. Ukita, “Deep back-projection networks for super-resolution,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 1664– 1673

    2018

  10. [10]

    Deep face super-resolution with iterative collaboration between attentive recovery and landmark estimation,

    C. Ma, Z. Jiang, Y . Rao, J. Lu, and J. Zhou, “Deep face super-resolution with iterative collaboration between attentive recovery and landmark estimation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 5569–5578

    2020

  11. [11]

    Feedback network for image super-resolution,

    Z. Li, J. Yang, Z. Liu, X. Yang, G. Jeon, and W. Wu, “Feedback network for image super-resolution,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 3867– 3876

    2019

  12. [12]

    Image super-resolution using deep convolutional networks,

    C. Dong, C. C. Loy, K. He, and X. Tang, “Image super-resolution using deep convolutional networks,”IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), vol. 38, no. 2, pp. 295–307, 2016

    2016

  13. [13]

    Enhanced deep residual networks for single image super-resolution,

    B. Lim, S. Son, H. Kim, S. Nah, and K. M. Lee, “Enhanced deep residual networks for single image super-resolution,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2017, pp. 136–144

    2017

  14. [14]

    Image super- resolution using very deep residual channel attention networks,

    Y . Zhang, K. Li, K. Li, L. Wang, B. Zhong, and Y . Fu, “Image super- resolution using very deep residual channel attention networks,” in European Conference on Computer Vision (ECCV), 2018, pp. 286–301

    2018

  15. [15]

    Swinir: Image restoration using swin transformer,

    J. Liang, J. Cao, G. Sun, K. Zhang, L. Van Gool, and R. Timofte, “Swinir: Image restoration using swin transformer,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, 2021, pp. 1833–1844

    2021

  16. [16]

    Generative adversarial nets,

    I. J. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio, “Generative adversarial nets,” in Advances in Neural Information Processing Systems (NeurIPS), 2014, pp. 2672–2680

    2014

  17. [17]

    Photo-realistic single image super-resolution using a generative adversarial network,

    C. Ledig, L. Theis, F. Husz ´ar, J. Caballero, A. Cunningham, A. Acosta, A. Aitken, A. Tejani, J. Totz, Z. Wang, and W. Shi, “Photo-realistic single image super-resolution using a generative adversarial network,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 4681–4690

    2017

  18. [18]

    Esrgan: Enhanced super-resolution generative adversarial networks,

    X. Wang, K. Yu, S. Wu, J. Gu, Y . Liu, C. Dong, Y . Qiao, and C. C. Loy, “Esrgan: Enhanced super-resolution generative adversarial networks,” in European Conference on Computer Vision (ECCV) Workshops, 2018, pp. 63–79

    2018

  19. [19]

    Super-fan: Integrated facial landmark localization and super-resolution of real-world low resolution faces in arbitrary poses with gans,

    A. Bulat and G. Tzimiropoulos, “Super-fan: Integrated facial landmark localization and super-resolution of real-world low resolution faces in arbitrary poses with gans,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 109–117

    2018

  20. [20]

    Fsrnet: End-to-end learning face super-resolution with facial priors,

    Y . Chen, Y . Tai, X. Liu, C. Shen, and J. Yang, “Fsrnet: End-to-end learning face super-resolution with facial priors,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 2492–2501

    2018

  21. [21]

    Vqfr: Blind face restoration with vector-quantized dictionary and parallel decoder,

    Y . Gu, X. Wang, L. Xie, C. Dong, G. Li, Y . Shan, and M.-M. Cheng, “Vqfr: Blind face restoration with vector-quantized dictionary and parallel decoder,” inEuropean Conference on Computer Vision (ECCV), 2022, pp. 126–143

    2022

  22. [22]

    Blind face restoration via deep multi-scale component dictionaries,

    X. Li, C. Chen, S. Zhou, X. Lin, W. Zuo, and L. Zhang, “Blind face restoration via deep multi-scale component dictionaries,” inEuropean Conference on Computer Vision (ECCV), 2020, pp. 399–415

    2020

  23. [23]

    Image super-resolution via iterative refinement,

    C. Saharia, J. Ho, W. Chan, T. Salimans, D. J. Fleet, and M. Norouzi, “Image super-resolution via iterative refinement,”IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), vol. 45, no. 4, pp. 4713–4726, 2023

    2023

  24. [24]

    Difface: Blind face restoration with diffused error contraction,

    Z. Yue and C. C. Loy, “Difface: Blind face restoration with diffused error contraction,”IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), vol. 46, no. 12, pp. 9991–10 004, 2024. 10

    2024

  25. [25]

    Inversion by direct iteration: An alter- native to denoising diffusion for image restoration,

    M. Delbracio and P. Milanfar, “Inversion by direct iteration: An alter- native to denoising diffusion for image restoration,”Transactions on Machine Learning Research (TMLR), 2023

    2023

  26. [26]

    Deep equilibrium models,

    S. Bai, J. Z. Kolter, and V . Koltun, “Deep equilibrium models,” in Advances in Neural Information Processing Systems (NeurIPS), 2019

    2019

  27. [27]

    Consistency models,

    Y . Song, P. Dhariwal, M. Chen, and I. Sutskever, “Consistency models,” inInternational Conference on Machine Learning (ICML), 2023, pp. 32 211–32 252

    2023

  28. [28]

    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,” inProceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 2223–2232

    2017

  29. [29]

    Refstar: Blind face image restoration with reference selection, transfer, and reconstruction,

    Z. Yin, J. Chen, M. Liu, Z. Wang, F. Li, R. Pei, X. Li, R. W. Lau, and W. Zuo, “Refstar: Blind face image restoration with reference selection, transfer, and reconstruction,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 40, no. 14, 2026, pp. 12 053–12 062

    2026

  30. [30]

    Towards unsupervised blind face restoration using diffusion prior,

    T. Kuai, S. Honari, I. Gilitschenski, and A. Levinshtein, “Towards unsupervised blind face restoration using diffusion prior,” inProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2025, pp. 1839–1849

    2025

  31. [31]

    The perception-distortion tradeoff,

    Y . Blau and T. Michaeli, “The perception-distortion tradeoff,” inPro- ceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 6228–6237

    2018

  32. [32]

    Searching for mobilenetv3,

    A. Howard, M. Sandler, G. Chu, L.-C. Chen, B. Chen, M. Tan, W. Wang, Y . Zhu, R. Pang, V . Vasudevan, Q. V . Le, and H. Adam, “Searching for mobilenetv3,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2019, pp. 1314–1324

    2019

  33. [33]

    Efficientnet: Rethinking model scaling for convolutional neural networks,

    M. Tan and Q. V . Le, “Efficientnet: Rethinking model scaling for convolutional neural networks,” inInternational Conference on Machine Learning (ICML), 2019, pp. 6105–6114

    2019

  34. [34]

    Quantization and training of neural networks for efficient integer-arithmetic-only inference,

    B. Jacob, S. Kligys, B. Chen, M. Zhu, M. Tang, A. Howard, H. Adam, and D. Kalenichenko, “Quantization and training of neural networks for efficient integer-arithmetic-only inference,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 2704–2713

    2018

  35. [35]

    Perceptual losses for real- time style transfer and super-resolution,

    J. Johnson, A. Alahi, and L. Fei-Fei, “Perceptual losses for real- time style transfer and super-resolution,” inEuropean Conference on Computer Vision (ECCV), 2016, pp. 694–711

    2016

  36. [36]

    Arcface: Additive angular margin loss for deep face recognition,

    J. Deng, J. Guo, N. Xue, and S. Zafeiriou, “Arcface: Additive angular margin loss for deep face recognition,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 4690–4699

    2019

  37. [37]

    A style-based generator architecture for generative adversarial networks,

    T. Karras, S. Laine, and T. Aila, “A style-based generator architecture for generative adversarial networks,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 4401–4410

    2019

  38. [38]

    Real-esrgan: Training real- world blind super-resolution with pure synthetic data,

    X. Wang, L. Xie, C. Dong, and Y . Shan, “Real-esrgan: Training real- world blind super-resolution with pure synthetic data,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, 2021, pp. 1905–1914

    2021

  39. [39]

    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 (CVPR), 2018, pp. 586–595

    2018

  40. [40]

    Image-to-image translation with conditional adversarial networks,

    P. Isola, J.-Y . Zhu, T. Zhou, and A. A. Efros, “Image-to-image translation with conditional adversarial networks,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 1125–1134

    2017

  41. [41]

    Spectral normal- ization for generative adversarial networks,

    T. Miyato, T. Kataoka, M. Koyama, and Y . Yoshida, “Spectral normal- ization for generative adversarial networks,” inInternational Conference on Learning Representations (ICLR), 2018. Fuchen LiFuchen Li is currently pursuing the Second M.S. degree in Electrical and Computer Engineering with the Herbert Wertheim College of Engineering, University of Florid...

    2018