Pith. sign in

REVIEW 4 major objections 5 minor 36 references

Uncertainty-Aware Regularization for Image-to-Image Translation

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read A total-variation penalty on the predicted error-shape map improves both translation quality and aleatoric uncertainty estimates in medical image-to-image translation, especially under noise and artifacts.

desk verdict The new WCE-FICE dataset is a real asset, but the reported λ makes the UAR term numerically inert, so the central claim is unproven. read the letter →

arxiv 2412.01705 v1 pith:RUMYFKSL submitted 2024-11-24 cs.CV cs.AIeess.IV

classification cs.CVcs.AIeess.IV
keywords uncertaintyquantificationaleatoricimage-to-imagetranslationtotalvariationregularizationgeneralizednormaldistributionmedicalimagingcapsuleendoscopygenerativeadversarialnetworks
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 aleatoric uncertainty maps in medical image-to-image translation can be made more truthful and useful by adding a total-variation penalty on the predicted shape parameter of the per-pixel error distribution. The penalty, called Uncertainty-Aware Regularization (UAR), is a single model-agnostic term in the generator loss, and the paper's experiments compare a conditional GAN trained with and without it on two medical imaging datasets. The reported result is that UAR consistently improves reconstruction metrics (LPIPS and RRMSE, with comparable or better SSIM/PSNR) and produces uncertainty maps that are less noisy, more aligned with image structure, and better at flagging injected artifacts as unfamiliar territory. If these results hold, the practical payoff is that downstream users of medical translation systems could read predicted uncertainty maps as signals of where the model lacks knowledge, rather than as reflections of pixel-level noise.

What carries the argument

The load-bearing object is the generalized normal distribution (GND) placed on per-pixel reconstruction residuals, with the shape parameter $\beta$ controlling tail weight and the scale $\alpha$ controlling spread; aleatoric uncertainty is the variance $\alpha^2 \Gamma(3/\beta)/\Gamma(1/\beta)$. The paper's mechanism is to add the total-variation penalty $R_{\beta_i}$ on the predicted $\beta$ map to the generator loss, under the prior that residuals of good reconstructions are piece-wise continuous. Total variation preserves edges, so the penalty suppresses spurious high-frequency noise in $\beta$ without flattening genuine uncertainty boundaries; the ablation shows the squared-gradient (L2) variant smooths edges more, while isotropic total variation gives the best reported metrics on the WCE dataset.

What would settle it

Use a synthetic paired image set whose true per-pixel residual variance is known and changes discontinuously at sharp boundaries, then train the baseline and UAR variants and compare predicted uncertainty at those boundaries: if UAR systematically flattens known genuine variance discontinuities, the piecewise-continuity prior is removing real signal, not just spurious noise.

Watch

Extended reading notes

Core claim

The central discovery is that imposing a simple spatial smoothness prior on the distribution-shape parameter $\beta$ of the assumed generalized normal residual distribution yields better aleatoric uncertainty estimates and better reconstructions than training the same conditional GAN without the prior. The generator outputs per-pixel $\alpha$ and $\beta$ alongside the translated image; UAR adds a regularized, edge-preserving total-variation penalty on the predicted $\beta$ map to the generator loss. In experiments with Gaussian, uniform and impulse noise, UAR consistently lowers LPIPS and RRMSE relative to the non-regularized baseline, with comparable or better SSIM/PSNR, and its uncertainty maps concentrate on genuinely uncertain structures rather than spreading uniformly. With injected circular and ring artifacts, UAR marks the artifact region as high-uncertainty with sharp boundaries, while the baseline fails to distinguish familiar from novel regions.

Load-bearing premise

The load-bearing premise is that good reconstructions have piece-wise continuous pixel residuals, so neighbouring pixels' predicted error-shape values should be similar; if genuine uncertainty can change sharply between adjacent pixels independently of image content, the regularizer will smooth away real signal.

Editorial extensions

If this is right

  • On the two tested medical datasets, UAR lowers LPIPS and RRMSE across Gaussian, uniform and impulse noise levels while keeping SSIM/PSNR comparable or better.
  • Uncertainty maps trained with UAR concentrate on genuinely difficult structures and injected artifacts, with sharp edges, whereas the non-regularized baseline spreads high uncertainty broadly.
  • Because UAR is a lightweight penalty on $\beta$, it can be plugged into existing conditional GAN translation models without sequential uncertainty-estimation stages or extra forward passes.
  • The ablation indicates the isotropic total-variation version balances edge preservation and noise suppression better than an L2 gradient penalty, which smooths away uncertainty boundaries.

Reading between the lines

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

  • A natural next test is whether penalizing the scale $\alpha$ or the full variance map, instead of only the shape $\beta$, changes the noise/edge trade-off; that would show whether $\beta$ is the essential channel or just a convenient one.
  • The artifact experiments suggest a practical side effect the paper does not claim: the regularized $\beta$ map could serve as a cheap out-of-distribution signal, since unfamiliar structures stand out as compact high-uncertainty regions.
  • The new paired WCE-to-FICE dataset opens a benchmark for calibrating uncertainty maps with known injected noise variances, which would let future work quantify whether UAR's sharper maps are also better calibrated.
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

4 major / 5 minor

Summary. The paper proposes an Uncertainty-Aware Regularization (UAR) term for image-to-image translation. The method models per-pixel residuals with a generalized normal distribution, predicting shape (β) and scale (α) parameters, and adds a total-variation penalty on the predicted β map to the generator loss. Experiments on a new wireless capsule endoscopy dataset and a public colonoscopy dataset are conducted under synthetic Gaussian, uniform, and impulse noise, as well as injected circular/ring artifacts. The authors report improved LPIPS and RRMSE reconstruction metrics, and qualitatively smoother, more semantically structured aleatoric uncertainty maps compared to a baseline without the regularizer. The paper also introduces a new paired WCE-to-FICE dataset.

Significance. If the central claims hold, the method is simple, model-agnostic, and computationally cheap, and the new dataset is a useful community resource. The likelihood algebra in Eqs. (1)-(6) is correct, and the regularizer is clearly defined. However, the paper's main claim of 'better uncertainty estimations' rests almost entirely on visual inspection of uncertainty maps; no quantitative uncertainty metric is reported. More seriously, the reported regularization weight λ=10^-12 appears numerically negligible, which would make the UAR term unable to influence optimization, contradicting the reported differences between baseline and UAR. The ablation in Section 5 also contains an internal inconsistency about the effective λ. These issues are load-bearing for the attribution of the observed improvements to UAR and for the validity of the uncertainty-quality claims.

major comments (4)
  1. [§3.3, Eq. (11); §5 Ablation II] The reported λ=10^-12 makes the UAR term λRβi roughly six orders of magnitude smaller than the NLL term (for 490×490 images with O(1) β gradients, λRβi ≈ 10^-7 per image versus wnll·Lnll ≈ 24 per image, giving a gradient ratio of about 10^-8 per pixel). Under these scales the regularizer cannot influence training, so the visibly different uncertainty maps and the improved metrics in Tables 1-3 cannot be attributed to Eq. (9). The ablation text compounds the inconsistency: λ=10^-7 is said to 'strike a balance,' while λ=10^-12 'as employed in this study' is said to yield satisfactory results, with the optimum anticipated in [10^-7, 10^-12]. The paper must clarify which λ actually produced the main results, re-run the experiments with the stated configuration, and report those numbers; if λ=10^-12 was truly used, the results should be statistically indistinguishable from the λ=0 baseline, which they are not.
  2. [§4, §4.1] There is no quantitative evaluation of the predicted uncertainty maps. The central claim that UAR provides 'better uncertainty estimations' is supported only by qualitative comparisons in Figures 4-7. The paper should report quantitative uncertainty metrics, for example calibration of predictive intervals, correlation between predicted uncertainty and per-pixel residual magnitude, or detection/localization metrics for the injected artifacts against known ground-truth masks. Without such a metric, the claim that UAR improves uncertainty estimation is not established, even if reconstruction quality improves.
  3. [§3.2, Fig. 2] Because UAR by construction penalizes spatial variation of β, observing smoother β/uncertainty maps under UAR is a direct consequence of the loss, not independent evidence that the resulting maps are more accurate. The 'benign assumption' that good reconstructions have piece-wise continuous residuals, so neighboring β values should be similar, is asserted but not tested. An independent test is needed, e.g., injecting noise or artifacts with known spatial support and measuring whether UAR's uncertainty maps better separate corrupted from uncorrupted regions against a ground-truth mask, compared with the baseline.
  4. [§5, Table 3] The statement 'Imposing these constraints does not negatively impact the reconstruction quality, as seen in Table 3' is contradicted by the UAR Aniso row, where LPIPS (0.133 vs 0.128) and RRMSE (0.215 vs 0.174) are worse than the baseline. This claim should be revised and the degradation caused by the anisotropic variant should be discussed.
minor comments (5)
  1. [§1] The introduction references 'Table 5' when discussing improved reconstruction quality, but the manuscript contains only Tables 1-3; the citation should be corrected.
  2. [§3.3] The text says 'all results are reported on a test-set of another 5,000 image pairs' for the WCE dataset and then states the CPC dataset was split 80:20; the distinction between the two datasets should be clearer, including the number of test pairs for CPC.
  3. [Figures 2-7] The figures are dense and the small text labels (e.g., 'σ2', 'x', 'β') are difficult to read; higher-resolution panels or larger fonts would improve reproducibility of the qualitative claims.
  4. [§3.3] The sentence 'using twin-titan RTX GPUs' appears to refer to NVIDIA Titan RTX GPUs; the exact hardware should be named correctly.
  5. [§3.2] The regularization constant ϵ=10^-7 in Eq. (9) is introduced but its effect on the loss scale or on the effective regularization strength is never discussed; a brief note would help readers interpret the magnitude of λ.

Circularity Check

1 steps flagged · score 3.0 of 10

Qualitative uncertainty evaluation is partly self-confirming: TV on beta directly enforces smoother maps; quantitative reconstruction gains remain independent.

  1. self definitional [Section 3.2 (Eq. 9) and Section 4 (Fig. 5 discussion)]
    "We propose to suppress this spurious component for a more accurate estimation of uncertainty by penalizing large differences in the predicted residual distributions for neighboring pixels. ... As seen in columns 4 and 7 ( σ2), UAR generates less noisy uncertainty maps, consistent with the distinctive features within the images."

    The UAR term Rβi (Eq. 9) is defined as the total variation of the predicted β map, so minimizing it directly penalizes |∇β| between neighboring pixels. The paper's qualitative evidence that UAR yields 'less noisy uncertainty maps' therefore measures exactly the quantity the regularizer was designed to reduce. Because no ground-truth aleatoric uncertainty map is used, 'better uncertainty estimation' is partially equated with the regularizer's own objective, making the qualitative evaluation self-confirming by construction. The quantitative LPIPS/RRMSE/SSIM/PSNR improvements are independent and prevent the circularity from being total.

full rationale

The paper's quantitative derivation is mostly self-contained: Eq. 11 adds a total-variation penalty on the predicted β map to a standard GAN+NLL objective, and the reported improvements in LPIPS, RRMSE, SSIM, and PSNR are independent measurements against held-out test sets. Those reconstruction metrics are not circular. The only self-confirming component is the qualitative evaluation of the uncertainty maps: the regularizer is defined as the total variation of β, and the paper's evidence that UAR produces 'less noisy uncertainty maps' is a direct observation of the quantity being penalized. Since no ground-truth uncertainty is available, 'less noisy' and 'more coherent' coincide with the regularizer's objective rather than an independently measured property of uncertainty quality. This makes the qualitative uncertainty-improvement claim partly circular by construction, though the reconstruction gains provide indirect but independent support for the method's practical value. The reported λ = 10^-12 creates an apparent scale inconsistency — at that weight the penalty gradient is orders of magnitude smaller than the NLL gradient, yet visible differences are reported — but this is a correctness concern, not a circularity. Self-citations in the reference list are not load-bearing. Overall score 3.

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

No new physical or architectural entities are introduced. The central claim rests on the piece-wise continuity prior (domain assumption), the GND likelihood from prior work, and a set of hand-chosen loss weights. The UAR term is a standard TV penalty applied to a new target.

free parameters (5)
  • Regularization weight λ = 10^-12
    Chosen by hand after experimenting with 10^-12, 10^-7 and 10^-4 (§5 Ablation II). At 10^-12 the penalty is mathematically negligible relative to other loss terms, making the reported effect suspicious.
  • NLL loss weight w_nll = 10^-4
    Set in §3.3. Controls the contribution of the generalized-normal likelihood that trains α and β; the paper does not ablate this value.
  • Adversarial loss weight w_adv = 10^-3
    Set in §3.3. Standard for cGAN balance; not justified by analysis.
  • TV smoothing constant ε = 10^-7
    Added in Eq. 9 to avoid derivative singularity; value chosen without sensitivity analysis.
  • Activation epoch for UAR = epoch 5
    TV penalty is activated around epoch 5 to let α and β initialize; chosen heuristically in §3.3.
assumptions (5)
  • domain assumption Pixel residuals of good reconstructions are piece-wise continuous, so neighboring β values should be similar.
    Introduced as the 'benign assumption' in §3.2; this is the load-bearing prior that makes the TV penalty sensible.
  • domain assumption The generalized normal distribution (GND) with zero mean is an appropriate model for residuals.
    Adopted from [29] in §3.1; the paper does not compare with alternative residual distributions.
  • domain assumption Aleatoric uncertainty is the variance α²Γ(3/β)/Γ(1/β) of the predicted GND.
    Standard result used in §3.1 to convert distribution parameters into an uncertainty map.
  • standard math The negative log-likelihood in Eq. 6 is the correct objective up to an additive constant.
    Derived from the GND density; the algebra is standard.
  • ad hoc to paper Test-time synthetic noise and artifacts are a valid proxy for out-of-distribution uncertainty.
    The paper evaluates 'novel' regions by injecting noise and synthetic artifacts (Section 4), but does not validate that this proxy corresponds to clinically relevant out-of-distribution inputs.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Uncertainty-Aware Regularization for Image-to-Image Translation." pith.science (2026). https://pith.science/paper/RUMYFKSL

@misc{pith2026241201705,
  author       = {Pith},
  title        = {Pith review of: Uncertainty-Aware Regularization for Image-to-Image Translation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RUMYFKSL}},
  note         = {Machine review of arXiv:2412.01705}
}
read the original abstract

The importance of quantifying uncertainty in deep networks has become paramount for reliable real-world applications. In this paper, we propose a method to improve uncertainty estimation in medical Image-to-Image (I2I) translation. Our model integrates aleatoric uncertainty and employs Uncertainty-Aware Regularization (UAR) inspired by simple priors to refine uncertainty estimates and enhance reconstruction quality. We show that by leveraging simple priors on parameters, our approach captures more robust uncertainty maps, effectively refining them to indicate precisely where the network encounters difficulties, while being less affected by noise. Our experiments demonstrate that UAR not only improves translation performance, but also provides better uncertainty estimations, particularly in the presence of noise and artifacts. We validate our approach using two medical imaging datasets, showcasing its effectiveness in maintaining high confidence in familiar regions while accurately identifying areas of uncertainty in novel/ambiguous scenarios.

Figures

Figures reproduced from arXiv: 2412.01705 by the authors.

Figure 1
Figure 1. Non-probabilistic image translation (green) optimizes point-estimates for the residuals between the predicted and the target [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The figure shows the shape β parameter and predicted aleatoric uncertainty of an image at different epochs during the training. Without regularization (first row), the variances in the predictions (uncertainty) remain relatively the same throughout training. In contrast, with regularization (second row), the predicted uncertainty gets progressively less noisy and more semantically refined over the course of the trai… view at source ↗
Figure 3
Figure 3. Uncertainty estimate is sensitive to small changes in in [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Impact of noise on uncertainty prediction and image reconstruction. The figure illustrates the impact of varying levels of Gaussian [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Qualitative Comparison. As can be seen from columns 3 [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: UAR distinctly identifies regions affected by the artifi [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 30 canonical work pages

  1. [1]

    A deep learning framework for quality assessment and restoration in video endoscopy

    Sharib Ali, Felix Zhou, Adam Bailey, Barbara Braden, James E East, Xin Lu, and Jens Rittscher. A deep learning framework for quality assessment and restoration in video endoscopy. Medical image analysis, 68:101900, 2021

  2. [2]

    Endol2h: deep super-resolution for capsule endoscopy

    Yasin Almalioglu, Kutsev Bengisu Ozyoruk, Abdulkadir Gokce, Kagan Incetan, Guliz Irem Gokceler, Muhammed Ali Simsek, Kivanc Ararat, Richard J Chen, Nicholas J Durr, Faisal Mahmood, et al. Endol2h: deep super-resolution for capsule endoscopy. IEEE Transactions on Medical Imaging, 39(12):4297–4309, 2020

  3. [3]

    Medical image translation using an edge-guided genera- tive adversarial network with global-to-local feature fusion

    Hamed Amini Amirkolaee and Hamid Amini Amirkolaee. Medical image translation using an edge-guided genera- tive adversarial network with global-to-local feature fusion. Journal of Biomedical Research, 36(6):409, 2022

  4. [4]

    Label-informed cardiac magnetic resonance image synthesis through condi- tional generative adversarial networks

    Sina Amirrajab, Yasmina Al Khalil, Cristian Lorenz, J ¨urgen Weese, Josien Pluim, and Marcel Breeuwer. Label-informed cardiac magnetic resonance image synthesis through condi- tional generative adversarial networks. Computerized Medi- cal Imaging and Graphics, 101:102123, 2022

  5. [5]

    Medgan: Medical image translation using gans

    Karim Armanious, Chenming Jiang, Marc Fischer, Thomas K¨ustner, Tobias Hepp, Konstantin Nikolaou, Sergios Ga- tidis, and Bin Yang. Medgan: Medical image translation using gans. Computerized medical imaging and graphics , 79:101684, 2020

  6. [6]

    Test-time data aug- mentation for estimation of heteroscedastic aleatoric uncer- tainty in deep neural networks

    Murat Seckin Ayhan and Philipp Berens. Test-time data aug- mentation for estimation of heteroscedastic aleatoric uncer- tainty in deep neural networks. In Medical Imaging with Deep Learning, 2018

  7. [7]

    Uncertainty estimation in contrast- enhanced mr image translation with multi-axis fusion

    Ivo M Baltruschat, Parvaneh Janbakhshi, Melanie Dohmen, and Matthias Lenga. Uncertainty estimation in contrast- enhanced mr image translation with multi-axis fusion. In 2024 IEEE International Symposium on Biomedical Imag- ing (ISBI), pages 1–5. IEEE, 2024

  8. [8]

    Image synthe- sis in multi-contrast mri with conditional generative adver- sarial networks

    Salman UH Dar, Mahmut Yurt, Levent Karacan, Aykut Erdem, Erkut Erdem, and Tolga Cukur. Image synthe- sis in multi-contrast mri with conditional generative adver- sarial networks. IEEE transactions on medical imaging , 38(10):2375–2388, 2019

Show all 36 references
  1. [9]

    Aleatory or epis- temic? does it matter? Structural safety, 31(2):105–112, 2009

    Armen Der Kiureghian and Ove Ditlevsen. Aleatory or epis- temic? does it matter? Structural safety, 31(2):105–112, 2009

  2. [10]

    Arsdm: colonoscopy images synthesis with adaptive refinement se- mantic diffusion models

    Yuhao Du, Yuncheng Jiang, Shuangyi Tan, Xusheng Wu, Qi Dou, Zhen Li, Guanbin Li, and Xiang Wan. Arsdm: colonoscopy images synthesis with adaptive refinement se- mantic diffusion models. In International conference on medical image computing and computer-assisted interven- tion...

  3. [11]

    Error reduction through post processing for wireless capsule endoscope video

    P ˚al Anders Floor, Ivar Farup, Marius Pedersen, and Øistein Hovde. Error reduction through post processing for wireless capsule endoscope video. EURASIP Journal on Image and Video Processing, 2020:1–15, 2020

  4. [12]

    Medsrgan: medical images super- resolution using generative adversarial networks

    Yuchong Gu, Zitao Zeng, Haibin Chen, Jun Wei, Yaqin Zhang, Binghui Chen, Yingqin Li, Yujuan Qin, Qing Xie, Zhuoren Jiang, et al. Medsrgan: medical images super- resolution using generative adversarial networks. Multime- dia Tools and Applications, 79:21815–21840, 2020

  5. [13]

    Google photos labeled black peo- ple’gorillas’

    Jessica Guynn. Google photos labeled black peo- ple’gorillas’. USA today, 1, 2015

  6. [14]

    Denoising dif- fusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020

  7. [15]

    Image-to-image translation with conditional adver- sarial networks

    Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, and Alexei A Efros. Image-to-image translation with conditional adver- sarial networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1125–1134, 2017

  8. [16]

    Diffusion models in medical imaging: A comprehensive survey

    Amirhossein Kazerouni, Ehsan Khodapanah Aghdam, Moein Heidari, Reza Azad, Mohsen Fayyaz, Ilker Haci- haliloglu, and Dorit Merhof. Diffusion models in medical imaging: A comprehensive survey. Medical Image Analysis, page 102846, 2023

  9. [17]

    What uncertainties do we need in bayesian deep learning for computer vision? Advances in neural information processing systems, 30, 2017

    Alex Kendall and Yarin Gal. What uncertainties do we need in bayesian deep learning for computer vision? Advances in neural information processing systems, 30, 2017

  10. [18]

    Precomputed real-time texture synthesis with markovian generative adversarial networks

    Chuan Li and Michael Wand. Precomputed real-time texture synthesis with markovian generative adversarial networks. In Computer Vision–ECCV 2016: 14th European Confer- ence, Amsterdam, The Netherlands, October 11-14, 2016, Proceedings, Part III 14, pages 702–716. Springer, 2016

  11. [19]

    Toward clinically assisted colorectal polyp recognition via structured cross-modal representation consistency

    Weijie Ma, Ye Zhu, Ruimao Zhang, Jie Yang, Yiwen Hu, Zhen Li, and Li Xiang. Toward clinically assisted colorectal polyp recognition via structured cross-modal representation consistency. In International Conference on Medical Image Computing and Computer-Assisted Intervention,...

  12. [20]

    Pe 16-007 technical report, u.s

    NHTSA. Pe 16-007 technical report, u.s. department of transportation, national highway traffic safety admin- istration. https://static.nhtsa.gov/odi/inv/ 2016/INCLA-PE16007-7876.PDF, Jan 2017. [Ac- cessed 22-Feb-2023]

  13. [21]

    Estimating the mean and variance of the target probability distribution

    David A Nix and Andreas S Weigend. Estimating the mean and variance of the target probability distribution. In Pro- ceedings of 1994 ieee international conference on neural networks (ICNN’94), volume 1, pages 55–60. IEEE, 1994

  14. [22]

    Image translation for medical image generation: Ischemic stroke lesion segmentation

    Moritz Platscher, Jonathan Zopes, and Christian Federau. Image translation for medical image generation: Ischemic stroke lesion segmentation. Biomedical Signal Processing and Control, 72:103283, 2022

  15. [23]

    Validating uncertainty in medical image translation

    Jacob C Reinhold, Yufan He, Shizhong Han, Yunqiang Chen, Dashan Gao, Junghoon Lee, Jerry L Prince, and Aaron Carass. Validating uncertainty in medical image translation. In 2020 IEEE 17th International Symposium on Biomedical Imaging (ISBI), pages 95–98. IEEE, 2020

  16. [24]

    U- net: Convolutional networks for biomedical image segmen- tation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U- net: Convolutional networks for biomedical image segmen- tation. In Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, pa...

  17. [25]

    Clinical utility of capsule endoscopy with 9884 flexible spectral imaging color enhancement for diagnosis of small bowel lesions

    Yasushi Sato, Tamotsu Sagawa, Masahiro Hirakawa, Hi- royuki Ohnuma, Takahiro Osuga, Yutaka Okagawa, Fu- mito Tamura, Hiroto Horiguchi, Kohichi Takada, Tsuyoshi Hayashi, et al. Clinical utility of capsule endoscopy with 9884 flexible spectral imaging color enhancement for diagn...

  18. [26]

    The european legal framework for medical ai

    David Schneeberger, Karl St ¨oger, and Andreas Holzinger. The european legal framework for medical ai. In Interna- tional Cross-Domain Conference for Machine Learning and Knowledge Extraction, pages 209–226. Springer, 2020

  19. [27]

    A generative adversarial network based super-resolution approach for capsule endoscopy images

    Mehmet Turan. A generative adversarial network based super-resolution approach for capsule endoscopy images. Medicine Science, 10(3):1002–1007, 2021

  20. [28]

    Ro- bustness via uncertainty-aware cycle consistency

    Uddeshya Upadhyay, Yanbei Chen, and Zeynep Akata. Ro- bustness via uncertainty-aware cycle consistency. Advances in neural information processing systems, 34:28261–28273, 2021

  21. [29]

    Uncertainty-guided progressive gans for medical image translation

    Uddeshya Upadhyay, Yanbei Chen, Tobias Hepp, Sergios Gatidis, and Zeynep Akata. Uncertainty-guided progressive gans for medical image translation. In Medical Image Com- puting and Computer Assisted Intervention–MICCAI 2021: 24th International Conference, Strasbourg, France, Se...

  22. [30]

    Uncertainty-aware gan with adaptive loss for robust mri image enhancement

    Uddeshya Upadhyay, Viswanath P Sudarshan, and Suyash P Awate. Uncertainty-aware gan with adaptive loss for robust mri image enhancement. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 3255– 3264, 2021

  23. [31]

    Bovik, H.R

    Zhou Wang, A.C. Bovik, H.R. Sheikh, and E.P. Simoncelli. Image quality assessment: from error visibility to struc- tural similarity. IEEE Transactions on Image Processing , 13(4):600–612, 2004

  24. [32]

    Mri cross-modality image-to- image translation

    Qianye Yang, Nannan Li, Zixu Zhao, Xingyu Fan, Eric I- Chao Chang, and Yan Xu. Mri cross-modality image-to- image translation. Scientific reports, 10(1):3753, 2020

  25. [33]

    Diana E Yung, John N Plevris, Romain Leenhardt, Xavier Dray, Anastasios Koulaouzidis, and ESGE Small Bowel Re- search Working Group. Poor quality of small bowel capsule endoscopy images has a significant negative effect in the di- agnosis of small bowel malignancy.Clinical and...

  26. [34]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shecht- man, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In Proceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 586–595, 2018

  27. [35]

    Translating and segmenting multimodal medical volumes with cycle-and shape-consistency generative adversarial network

    Zizhao Zhang, Lin Yang, and Yefeng Zheng. Translating and segmenting multimodal medical volumes with cycle-and shape-consistency generative adversarial network. In Pro- ceedings of the IEEE conference on computer vision and pat- tern Recognition, pages 9242–9251, 2018

  28. [36]

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

    Jun-Yan Zhu, Taesung Park, Phillip Isola, and Alexei A Efros. Unpaired image-to-image translation using cycle- consistent adversarial networks. In Proceedings of the IEEE international conference on computer vision , pages 2223– 2232, 2017. 9885

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.