Pith. sign in

REVIEW 4 major objections 6 minor 62 references

NTIRE 2026 Low-light Enhancement: Twilight Cowboy Challenge

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Ten teams beat a night-burst RAW baseline by up to 6.49 dB

desk verdict A useful new benchmark dataset for handheld low-light burst RAW restoration, but the ground-truth construction is not validated and casts doubt on the headline SOTA claims. read the letter →

arxiv 2608.09782 v1 pith:BR32WZTP submitted 2026-08-10 cs.CV

Aleksei Khalin , Egor Ershov , Artyom Panshin , Sergey Korchagin , Georgiy Lobarev , Arseniy Terekhin , Sofiia Dorogova , Amir Shamsutdinov
show 41 more authors
Yasin Mamedov Bakhtiyar Khalfin Bogdan Sheludko Emil Zilyaev Nikola Banić Georgy Perevozchikov Radu Timofte Shuai Liu Yuqian Zhang Lize Zhang Yibin Huang Chaoyu Feng Luyang Wang Xiaotao Wang Dongqing Zou Lei Lei Tianli Liu Dejun Hao Chunxia Lei Furkan K{i}nl{i} Andrei Mironov Alexander Dikov Aleksei Sadokhin Vladimir Zvorygin Constantine Habarlak Shuwei Yue Egor Mirantsov Daniil Okunev Dmitry Arkhipov Aleksandr Yugay Anas M. Ali Bilel Benjdira Wadii Boulila Wei Zhou Linfeng Li Lingdong Kong Jiachen Tu Guoyi Xu Yaoxin Jiang Jiajia Liu Yaokun Shi
This is my paper · ORCID
classification cs.CV
keywords low-lightenhancementburstdenoisingRAWimagerestorationhandheldmisalignmentbenchmarkdatasetBayerdomainsmartphonephotography
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 challenge report claims that a deliberately hard benchmark, merging five misaligned handheld RAW smartphone frames into one clean image, cleanly separates restoration methods, and that ten submitted systems beat the provided alignment-plus-BM3D baseline, the best by +6.49 dB PSNR and +0.0101 SSIM. The paper introduces a dataset of 585 real indoor and outdoor nighttime scenes, each paired with a noise-free ground truth built from 270 tripod frames, and a three-stage evaluation protocol that ranks systems by a weighted combination of PSNR and SSIM ranks. If the ground-truth construction is sound, the leaderboard establishes a credible reference point for joint motion compensation and denoising directly in the RAW/Bayer domain.

What carries the argument

The machinery is the ground-truth generation protocol: for each scene, 270 tripod RAW frames are intensity-aligned, frames whose global mean falls outside the fitted normal distribution's 99.7% interval are rejected, and the per-pixel normal-distribution mean of the survivors becomes the clean RAW reference, followed by quantile normalization. This reference is paired with five handheld frames selected every tenth frame from a handheld sequence, defining the input-target pairs. The evaluation machinery is the composite score $S=0.6\cdot R_P+0.4\cdot R_S$ over PSNR rank and SSIM rank on a private 85-pair test set, which converts raw quality numbers into a single ranking.

What would settle it

Re-generate the ground-truth references using a per-pixel median or trimmed mean instead of the normal-distribution mean, rerun the submitted pipelines on the same 85 test inputs, and check whether the leaderboard ordering changes; a substantial reordering would show the reported improvements are partly an artifact of the reference construction rather than of restoration quality.

Watch

Extended reading notes

Core claim

The central discovery is that the benchmark separates methods meaningfully: the best submission reaches 40.62 dB PSNR and 0.9875 SSIM against the baseline's 34.13 dB and 0.9774, a gap that the paper attributes to real progress on real-world noise, motion, and mixed illuminants. The winning approach combines three restoration backbones, optical-flow-guided deformable alignment, and 8-way test-time augmentation, at a high inference time of 514.6 s per sample, while several far lighter pipelines still exceed the baseline by wide margins. The paper argues that the dataset and three-stage protocol offer a fair, reusable reference for the field, with the private test set protecting the final ranking from overfitting.

Load-bearing premise

The clean reference image is obtained by averaging many tripod frames after discarding frames whose global brightness lies outside a fitted normal distribution, which assumes pixel noise is normally distributed and that all dynamic scene content is rejected; if the reference contains ghosting or bias, every PSNR and SSIM score in the table shifts.

Editorial extensions

If this is right

  • Methods that jointly align and denoise in the RAW/Bayer domain outperform pipelines that demosaic before fusion, since the top entries operate on Bayer data.
  • Efficient submissions (4.7–10.7 s per sample) stay within roughly 0.4–4.6 dB of the computationally heavy winner, suggesting practical mobile deployment is achievable without abandoning the benchmark's gains.
  • The baseline, detector-free matching plus homography and BM3D, acts as a floor that learned alignment and denoising consistently beat; future burst-RAW work can report directly against this table.
  • The +6.49 dB PSNR improvement quantifies the gap between simple averaging-plus-denoising and purpose-built burst restoration on realistic low-light scenes.

Reading between the lines

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

  • If the ground truth is biased by the 3-sigma rejection, for example if slow illumination changes inside the frame survive while moving lights outside the frame are removed, the benchmark may reward methods that suppress dynamic content; a useful test is to compare rankings on static-only subsets.
  • The near-saturated SSIM values (all above 0.975) suggest SSIM contributes little discrimination among top teams, so a metric more sensitive to texture or color could alter the composite ranking.
  • The dataset uses fixed capture settings (ISO 100, 1/10 s), covering one operating point; extending to varied exposures or sensor gains would test whether the ranking order persists.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. This paper reports the NTIRE 2026 Twilight Cowboy Challenge on burst-based low-light RAW enhancement. The organizers collected 585 real-world scenes, each with five misaligned handheld RAW frames and a ground-truth RAW image constructed by averaging 270 tripod-captured frames after outlier rejection. The evaluation uses a three-stage protocol with a private 85-pair test set, ranking participants by a weighted combination of PSNR and SSIM ranks. Ten teams surpassed the provided ASpanFormer+BM3D baseline, with the winning team MiAlgo reaching 40.62 dB PSNR and 0.9875 SSIM. The paper claims these results establish new state-of-the-art performance for burst-based low-light image enhancement.

Significance. If the ground-truth construction is valid, the dataset and protocol provide a useful public resource for handheld low-light RAW burst restoration, with the private test set and CodaBench evaluation reducing direct overfitting. The paper is transparent about the capture procedure, baseline, and leaderboard, and it makes the results publicly available. The main assessment hinges on whether the per-pixel averaging procedure in Section 2.2 produces clean, motion-free references; the manuscript provides no validation of this assumption, and the winning team's reported 25% training-pair removal for spatial misalignment (Section 3.2) directly indicates that references are not always well aligned. The 'new state-of-the-art' claim is also not supported by comparison with existing published burst low-light methods on established benchmarks.

major comments (4)
  1. [Section 2.2] The ground-truth construction assumes that, after discarding frames whose global mean intensity falls outside the 99.7% interval, per-pixel values across the remaining 270 frames follow a normal distribution whose mean is a clean reference. This assumption fails for dynamic content inside the field of view, such as moving headlights, pedestrians, or blinking signs, which are not filtered by a global intensity threshold. Since every PSNR and SSIM value in Table 1 is computed against these references, the paper must provide evidence of reference cleanliness, for example residual statistics, manual inspection counts, or comparison with a robust estimator such as per-pixel median. Without this, the benchmark's validity is not established.
  2. [Section 3.2] The winning team MiAlgo reports manually removing about 25% of training pairs because of 'visible spatial misalignment between burst input and ground truth.' This is direct evidence that the ground-truth images are not cleanly aligned to the burst frames for a substantial fraction of the data. The paper should quantify how many of the 85 test pairs were affected by similar misalignment or ghosting, and should report whether any test references were manually inspected. If the test references contain the same artifacts, methods that reproduce the ghosted average are rewarded and genuinely clean reconstructions are penalized, which would bias the entire leaderboard.
  3. [Table 1, Section 2.4] The leaderboard reports mean PSNR and SSIM over 85 test images without confidence intervals, per-scene variance, or significance tests. Differences such as MiAlgo at 40.62 dB versus DH ISP at 40.23 dB, or BAU-Vision at 38.01 dB versus AXIOM at 37.85 dB, may not be statistically meaningful on this sample size. The paper should report standard errors, paired significance tests, or at least per-scene score distributions so that the ranking and the claimed improvements over the baseline can be properly interpreted.
  4. [Abstract and Section 4] The claim of establishing 'new state-of-the-art performance for burst-based low-light image enhancement' is not supported by any comparison with existing published burst low-light methods on established datasets or metrics. The reported numbers are only relative to the challenge baseline on the organizers' own private test set. The claim should either be restricted to 'state-of-the-art within this challenge' or be supplemented by a comparison with prior methods under a common evaluation protocol.
minor comments (6)
  1. [Section 2.2] There are several typographical errors, including 'the a normal distribution' and 'performed at the pixel level. To do so, the a normal distribution was fitted,' which should be corrected.
  2. [Section 2.1] The phrase 'five RAW images captured with misaligned RAW images of the same low-light scene' is redundant and should be reworded for clarity.
  3. [Section 3.1] The baseline description states 'the averaged frame, with reduced noise, was subsequently rocessed using the BM3D algorithm,' where 'rocessed' should read 'processed.'
  4. [Figure 1] The figure caption and axis labels would benefit from clearer units; the reader must infer that the x-axis is log-scaled inference time and the y-axis is mean PSNR from the text.
  5. [Section 2.3] The paper says 'The initial stage participants were provided with 250 training pairs and 50 validation samples,' but does not state whether the validation samples had public or withheld ground truth; this should be clarified for reproducibility of the protocol.
  6. [Section 4] The discussion notes the inference-time trade-off but does not report the computational resources used for training the participant methods; a brief note on GPU hardware or training cost would improve comparability.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the ranking is an empirical measurement of independent participant systems on a private test set, and the ground-truth construction is a definitional benchmark choice rather than a fitted input renamed as a prediction.

full rationale

This paper is a challenge report. Its central outputs, the leaderboard, the PSNR/SSIM improvements, and the state-of-the-art label, are empirical measurements of independently submitted participant systems evaluated on a private 85-pair test set; they are not derived from the organizers' assumptions by any equation. The ground-truth construction in Section 2.2 (per-pixel normal-distribution mean of 270 tripod frames after discarding frames outside the 99.7% intensity interval, followed by quantile normalization in Eq. 1) defines the reference image used by all teams and by the baseline. That is a dataset-construction choice, not a parameter fitted to the test set and then reported as a prediction. The baseline's BM3D noise parameter is fixed at 3/2047 after selection and is used only to produce a baseline score, not to predict participant performance. There is no load-bearing self-citation, no imported uniqueness theorem, and no ansatz smuggled in via the authors' prior work; the only external dataset-construction citation, SIDD [1], is standard practice. The label state-of-the-art for burst-based low-light image enhancement is meaningful only relative to this newly created benchmark, which is the normal definitional scope of a challenge report, not a circular derivation. A genuine validity concern appears in Section 3.2: team MiAlgo reports manually removing about 25% of training pairs because of visible spatial misalignment between burst input and ground truth, indicating that some reference images are not perfectly aligned or clean, and all reported PSNR and SSIM values inherit whatever bias the references contain. This is a benchmark-quality and correctness threat, not circularity: the test set was private, no solution is a restatement of the ground-truth averaging procedure, and the ranking is not produced by fitting the evaluation set. The score of 2 reflects the mild self-referential scope of the SOTA claim and the ground-truth-definitional caveat, not a reduction of the paper's results to its inputs.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The evaluation and the leaderboard depend on several hand-set thresholds and design choices: the baseline's BM3D noise parameter, the quantile normalization percentiles, the 3-sigma rejection threshold, and the ranking weights. These are not derived from independent benchmarks and should be scrutinized when interpreting the absolute scores and rankings.

free parameters (4)
  • Baseline BM3D noise level = sigma = 3/2047
    Section 3.1: the noise parameter is selected to maximize PSNR, then fixed. This tunes the baseline to the data and affects the reported improvement margins.
  • Ranking weights = 0.6 (PSNR), 0.4 (SSIM)
    Equation (2): the final score S is a hand-chosen weighted combination of PSNR and SSIM ranks. Different weights would change the leaderboard order.
  • Quantile normalization percentiles = 1st and 99th
    Equation (1): all images, including ground truth and outputs, are linearly scaled using the 1st and 99th percentiles. This choice affects the PSNR/SSIM values.
  • Ground-truth outlier rejection threshold = 99.7% (3-sigma)
    Section 2.2: frames whose average intensity falls outside the 99.7% interval are discarded. This hand-set threshold influences which frames contribute to the ground truth.
assumptions (4)
  • domain assumption The pixel-wise robust mean of 270 intensity-aligned frames after 3-sigma global rejection is a clean approximation of a stationary-camera ground truth.
    Section 2.2. The entire evaluation relies on this GT; dynamic objects or non-Gaussian noise could bias the mean and affect all PSNR/SSIM scores.
  • domain assumption PSNR and SSIM computed on quantile-normalized images are adequate proxies for perceptual quality in this challenge.
    Section 2.4 and Eq. 1. The ranking uses these metrics without validation against perceptual judges.
  • domain assumption The five frames selected as every tenth frame from a handheld sequence are representative of real burst photography misalignment.
    Section 2.1. The dataset's motion statistics depend on the specific handheld sequence and frame selection rule.
  • domain assumption The scene is static between the tripod ground-truth capture and the handheld capture, apart from global camera motion.
    Section 2.1. If dynamic objects appear differently between captures, the training pairs contain misregistration that methods must learn to ignore.

how reviews work

0 comments
Cite this review

Pith. "Pith review of NTIRE 2026 Low-light Enhancement: Twilight Cowboy Challenge." pith.science (2026). https://pith.science/paper/BR32WZTP

@misc{pith2026260809782,
  author       = {Pith},
  title        = {Pith review of: NTIRE 2026 Low-light Enhancement: Twilight Cowboy Challenge},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BR32WZTP}},
  note         = {Machine review of arXiv:2608.09782}
}
read the original abstract

This paper presents a review of the NTIRE 2026 Low-light Enhancement: Twilight Cowboy Challenge. The objective of the competition was to merge a set of misaligned smartphone images in the raw domain, captured in low-light conditions, into a single, clean image. Introduced setup simultaneously addresses two problems of low-light photography: visual degradations such as high noise and mixed scene illuminants, and the geometric inconsistencies caused by hand movement during multi-frame capture. To advance research in low-light and nighttime computational photography, a challenging dataset was collected comprising 585 real-world scenes, spanning indoor low-light and outdoor nighttime conditions, for training and benchmarking participant solutions. The competition employed a three-stage evaluation protocol: automatic validation via the CodaBench platform in stages one and two, followed by blind assessment on a private test set for the final ranking. Ten teams surpassed the established baseline, achieving improvements of up to +6.49 dB in PSNR and +0.0101 in SSIM, thereby establishing new state-of-the-art performance for burst-based low-light image enhancement. These results demonstrate significant progress in handling real-world noise, motion, and illumination variability in the low-light setting. Comprehensive results, leaderboards, and additional information are publicly available at https://nightimaging.org.

Figures

Figures reproduced from arXiv: 2608.09782 by the authors.

Figure 1
Figure 1. Comparison of mean PSNR scores and inference times [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Visual comparison of denoising approaches submitted to the final stage of the competition. Top-performing solutions excel at [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Testing pipeline of Avengers Assemble (MiAlgo). [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Overall Pipeline of “High-Precision Noise Transfer and [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: RestoRoBurr inference pipeline (html5attention3). [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: DUSKAN architecture (PSU). A learned per-stage logit α blends global spectral-spatial features (Path A) with polynomial￾basis KAN activations (Path B) inside a symmetric 4-level U-Net. 3.10. FengFans Our solution addresses multi-frame RAW denoising through a two-stage …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

62 extracted references · 19 canonical work pages

  1. [1]

    Abdelrahman Abdelhamed, Stephen Lin, and Michael S. Brown. A high-quality denoising dataset for smartphone cameras. InIEEE Conference on Computer Vision and Pat- tern Recognition (CVPR), 2018. 2

    work page 2018

  2. [2]

    NT-HAZE: A Benchmark Dataset for Re- alistic Night-time Image Dehazing

    Radu Ancuti, Codruta Ancuti, Radu Timofte, and Cos- min Ancuti. NT-HAZE: A Benchmark Dataset for Re- alistic Night-time Image Dehazing . InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2026. 2

    2026

  3. [3]

    NTIRE 2026 Nighttime Image Dehazing Challenge Report

    Radu Ancuti, Alexandru Brateanu, Florin Vasluianu, Raul Balmez, Ciprian Orhei, Codruta Ancuti, Radu Timofte, Cos- min Ancuti, et al. NTIRE 2026 Nighttime Image Dehazing Challenge Report . InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR) Workshops, 2026. 2

    2026

  4. [4]

    NTIRE 2026 Challenge on Single Image Re- flection Removal in the Wild: Datasets, Results, and Meth- ods

    Jie Cai, Kangning Yang, Zhiyuan Li, Florin Vasluianu, Radu Timofte, et al. NTIRE 2026 Challenge on Single Image Re- flection Removal in the Wild: Datasets, Results, and Meth- ods . InProceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR) Workshops,

    2026

  5. [5]

    Aspanformer: Detector-free image matching with adaptive span transformer

    Hongkai Chen, Zixin Luo, Lei Zhou, Yurun Tian, Ming- min Zhen, Tian Fang, David Mckinnon, Yanghai Tsin, and Long Quan. Aspanformer: Detector-free image matching with adaptive span transformer. InEuropean conference on computer vision, pages 20–36. Springer, 2022. 3, 5, 6

    work page 2022

  6. [6]

    Simple baselines for image restoration

    Liangyu Chen, Xiaojie Chu, Xiangyu Zhang, and Jian Sun. Simple baselines for image restoration. InEuropean confer- ence on computer vision, pages 17–33. Springer, 2022. 4, 5, 6, 7

    work page 2022

  7. [7]

    The Fourth Challenge on Image Super-Resolution (×4) at NTIRE 2026: Benchmark Results and Method Overview

    Zheng Chen, Kai Liu, Jingkai Wang, Xianglong Yan, Jianze Li, Ziqing Zhang, Jue Gong, Jiatong Li, Lei Sun, Xi- aoyang Liu, Radu Timofte, Yulun Zhang, et al. The Fourth Challenge on Image Super-Resolution (×4) at NTIRE 2026: Benchmark Results and Method Overview . InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR) W...

    2026

  8. [8]

    Low Light Image Enhancement Challenge at NTIRE 2026

    George Ciubotariu, Sharif S M A, Abdur Rehman, Fayaz Ali Dharejo, Rizwan Ali Naqvi, Marcos Conde, Radu Tim- ofte, et al. Low Light Image Enhancement Challenge at NTIRE 2026 . InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Work- shops, 2026. 2

    2026

Show all 62 references
  1. [9]

    High FPS Video Frame Interpolation Challenge at NTIRE 2026

    George Ciubotariu, Zhuyun Zhou, Yeying Jin, Zongwei Wu, Radu Timofte, et al. High FPS Video Frame Interpolation Challenge at NTIRE 2026 . InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2026. 2

    2026

  2. [10]

    Image denoising by sparse 3-D transform- domain collaborative filtering.IEEE Transactions on Image Processing, 16(8):2080–2095, 2007

    Kostadin Dabov, Alessandro Foi, Vladimir Katkovnik, and Karen Egiazarian. Image denoising by sparse 3-D transform- domain collaborative filtering.IEEE Transactions on Image Processing, 16(8):2080–2095, 2007. 3, 7

    2007

  3. [11]

    Bm3d image denoising with shape- adaptive principal component analysis

    Kostadin Dabov, Alessandro Foi, Vladimir Katkovnik, and Karen Egiazarian. Bm3d image denoising with shape- adaptive principal component analysis. InSPARS’09-Signal Processing with Adaptive Sparse Structured Representa- tions, 2009. 5

    2009

  4. [12]

    Superpoint: Self-supervised interest point detection and description

    Daniel DeTone, Tomasz Malisiewicz, and Andrew Rabi- novich. Superpoint: Self-supervised interest point detection and description. InProceedings of the IEEE conference on computer vision and pattern recognition workshops, pages 224–236, 2018. 7

    2018

  5. [13]

    D. L. Donoho and I. M. Johnstone. Ideal spatial adaptation by wavelet shrinkage.Biometrika, 81(3):425–455, 1994. 7

    1994

  6. [14]

    Burstormer: Burst image restoration and enhancement transformer

    Akshay Dudhane, Syed Waqas Zamir, Salman Khan, Fa- had Shahbaz Khan, and Ming-Hsuan Yang. Burstormer: Burst image restoration and enhancement transformer. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023, Vancouver, BC, Canada, June 17- 24, 2023, pa...

    2023

  7. [15]

    NTIRE 2026 Rip Current Detection and Segmentation (RipDetSeg) Chal- lenge Report

    Andrei Dumitriu, Aakash Ralhan, Florin Miron, Florin Ta- tui, Radu Tudor Ionescu, Radu Timofte, et al. NTIRE 2026 Rip Current Detection and Segmentation (RipDetSeg) Chal- lenge Report . InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) ...

    2026

  8. [16]

    Roma v2: Harder better faster denser feature matching

    Johan Edstedt, David Nordstr ¨om, Yushan Zhang, Georg B¨okman, Jonathan Astermark, Viktor Larsson, Anders Hey- den, Fredrik Kahl, M ˚arten Wadenb¨ack, and Michael Fels- berg. Roma v2: Harder better faster denser feature matching. CoRR, abs/2511.15706, 2025. 6

    2025 arXiv

  9. [17]

    Conde, Zongwei Wu, Yeying Jin, Radu Timofte, et al

    Omar Elezabi, Marcos V . Conde, Zongwei Wu, Yeying Jin, Radu Timofte, et al. Photography Retouching Trans- fer, NTIRE 2026 Challenge: Report . InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2026. 2

    2026

  10. [18]

    Greedy function approximation: a gra- dient boosting machine.Annals of statistics, pages 1189– 1232, 2001

    Jerome H Friedman. Greedy function approximation: a gra- dient boosting machine.Annals of statistics, pages 1189– 1232, 2001. 5

    2001

  11. [19]

    Gonzalez and Richard E

    Rafael C. Gonzalez and Richard E. Woods.Digital Image Processing. Pearson, 4th edition, 2018. 3

    2018

  12. [20]

    NTIRE 2026 Challenge on End-to-End Financial Receipt Restoration and Reasoning from Degraded Images: Datasets, Methods and Results

    Bochen Guan, Jinlong Li, Kangning Yang, Chuang Ke, Jie Cai, Florin Vasluianu, Radu Timofte, et al. NTIRE 2026 Challenge on End-to-End Financial Receipt Restoration and Reasoning from Degraded Images: Datasets, Methods and Results . InProceedings of the IEEE/CVF Conference on C...

    2026

  13. [21]

    NTIRE 2026 The 3rd Restore Any Image Model (RAIM) Challenge: AI Flash Portrait (Track 3)

    Ya-nan Guan, Shaonan Zhang, Hang Guo, Yawen Wang, Xinying Fan, Jie Liang, Hui Zeng, Guanyi Qin, Lishen Qu, Tao Dai, Shu-Tao Xia, Lei Zhang, Radu Timofte, et al. NTIRE 2026 The 3rd Restore Any Image Model (RAIM) Challenge: AI Flash Portrait (Track 3) . InProceedings of the IEEE...

    2026

  14. [22]

    NTIRE 2026 Challenge on Robust AI-Generated Image Detection in the Wild

    Aleksandr Gushchin, Khaled Abud, Ekaterina Shumitskaya, Artem Filippov, Georgii Bychkov, Sergey Lavrushkin, Mikhail Erofeev, Anastasia Antsiferova, Changsheng Chen, Shunquan Tan, Radu Timofte, Dmitriy Vatolin, et al. NTIRE 2026 Challenge on Robust AI-Generated Image Detection ...

    2026

  15. [23]

    Robust Deepfake De- tection, NTIRE 2026 Challenge: Report

    Benedikt Hopf, Radu Timofte, et al. Robust Deepfake De- tection, NTIRE 2026 Challenge: Report . InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2026. 2

    2026

  16. [24]

    Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen

    Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. Lora: Low-rank adaptation of large language models, 2021. 6

    2021

  17. [25]

    Averaging weights leads to wider optima and better generalization.arXiv preprint arXiv:1803.05407, 2018

    Pavel Izmailov, Dmitrii Podoprikhin, Timur Garipov, Dmitry Vetrov, and Andrew Gordon Wilson. Averaging weights leads to wider optima and better generalization.arXiv preprint arXiv:1803.05407, 2018. 7

    2018 arXiv

  18. [26]

    The First Challenge on Mobile Real-World Image Super- Resolution at NTIRE 2026: Benchmark Results and Method Overview

    Jiatong Li, Zheng Chen, Kai Liu, Jingkai Wang, Zihan Zhou, Xiaoyang Liu, Libo Zhu, Radu Timofte, Yulun Zhang, et al. The First Challenge on Mobile Real-World Image Super- Resolution at NTIRE 2026: Benchmark Results and Method Overview . InProceedings of the IEEE/CVF Conference...

    2026

  19. [27]

    NTIRE 2026 Challenge on Short-form UGC Video Restoration in the Wild with Generative Models: Datasets, Methods and Results

    Xin Li, Jiachao Gong, Xijun Wang, Shiyao Xiong, Bingchen Li, Suhang Yao, Chao Zhou, Zhibo Chen, Radu Timofte, et al. NTIRE 2026 Challenge on Short-form UGC Video Restoration in the Wild with Generative Models: Datasets, Methods and Results . InProceedings of the IEEE/CVF Con- ...

    2026

  20. [28]

    NTIRE 2026 The Second Challenge on Day and Night Raindrop Removal for Dual-Focused Images: Methods and Results

    Xin Li, Yeying Jin, Suhang Yao, Beibei Lin, Zhaoxin Fan, Wending Yan, Xin Jin, Zongwei Wu, Bingchen Li, Peishu Shi, Yufei Yang, Yu Li, Zhibo Chen, Bihan Wen, Robby Tan, Radu Timofte, et al. NTIRE 2026 The Second Challenge on Day and Night Raindrop Removal for Dual-Focused Imag...

    2026

  21. [29]

    Lightglue: Local feature matching at light speed

    Philipp Lindenberger, Paul-Edouard Sarlin, and Marc Polle- feys. Lightglue: Local feature matching at light speed. In Proceedings of the IEEE/CVF international conference on computer vision, pages 17627–17638, 2023. 7

    2023

  22. [30]

    The First Chal- lenge on Remote Sensing Infrared Image Super-Resolution at NTIRE 2026: Benchmark Results and Method Overview

    Kai Liu, Haoyang Yue, Zeli Lin, Zheng Chen, Jingkai Wang, Jue Gong, Radu Timofte, Yulun Zhang, et al. The First Chal- lenge on Remote Sensing Infrared Image Super-Resolution at NTIRE 2026: Benchmark Results and Method Overview . InProceedings of the IEEE/CVF Conference on Comp...

    2026

  23. [31]

    Conde, et al

    Shuhong Liu, Ziteng Cui, Chenyu Bao, Xuangeng Chu, Lin Gu, Bin Ren, Radu Timofte, Marcos V . Conde, et al. 3D Restoration and Reconstruction in Adverse Conditions: Re- alX3D Challenge Results . InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition...

    2026

  24. [32]

    NTIRE 2026 X- AIGC Quality Assessment Challenge: Methods and Results

    Xiaohong Liu, Xiongkuo Min, Guangtao Zhai, Qiang Hu, Jiezhang Cao, Yu Zhou, Wei Sun, Farong Wen, Zitong Xu, Yingjie Zhou, Huiyu Duan, Lu Liu, Jiarui Wang, Siqi Luo, Chunyi Li, Li Xu, Zicheng Zhang, Yue Shi, Yubo Wang, Minghong Zhang, Chunchao Guo, Zhichao Hu, Mingtao Chen, Xie...

    2026

  25. [33]

    An iterative image reg- istration technique with an application to stereo vision

    Bruce D Lucas and Takeo Kanade. An iterative image reg- istration technique with an application to stereo vision. In IJCAI’81: 7th international joint conference on Artificial in- telligence, pages 674–679, 1981. 5

    1981

  26. [34]

    Burst denoising with kernel prediction networks

    Ben Mildenhall, Jonathan T Barron, Jiawen Chen, Dillon Sharlet, Ren Ng, and Robert Carroll. Burst denoising with kernel prediction networks. InProceedings of the IEEE con- ference on computer vision and pattern recognition, pages 2502–2510, 2018. 6

    2018

  27. [35]

    NTIRE 2026 Challenge on Video Saliency Predic- tion: Methods and Results

    Andrey Moskalenko, Alexey Bryncev, Ivan Kosmynin, Kira Shilovskaya, Mikhail Erofeev, Dmitry Vatolin, Radu Timo- fte, et al. NTIRE 2026 Challenge on Video Saliency Predic- tion: Methods and Results . InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogn...

    2026

  28. [36]

    NTIRE 2026 Challenge on Efficient Burst HDR and Restoration: Datasets, Methods, and Results

    Hyunhee Park, Eunpil Park, Sangmin Lee, Radu Timofte, et al. NTIRE 2026 Challenge on Efficient Burst HDR and Restoration: Datasets, Methods, and Results . InProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2026. 2

    2026

  29. [37]

    NTIRE 2026 Challenge on Learned Smartphone ISP with Unpaired Data: Methods and Results

    Georgy Perevozchikov, Daniil Vladimirov, Radu Timofte, et al. NTIRE 2026 Challenge on Learned Smartphone ISP with Unpaired Data: Methods and Results . InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR) Workshops, 2026. 2

    2026

  30. [38]

    NTIRE 2026 The 3rd Restore Any Image Model (RAIM) Challenge: Professional Image Quality Assessment (Track 1)

    Guanyi Qin, Jie Liang, Bingbing Zhang, Lishen Qu, Ya-nan Guan, Hui Zeng, Lei Zhang, Radu Timofte, et al. NTIRE 2026 The 3rd Restore Any Image Model (RAIM) Challenge: Professional Image Quality Assessment (Track 1) . InPro- ceedings of the IEEE/CVF Conference on Computer Vision...

    2026

  31. [39]

    Recursive multi- exposure alignment with spatiotemporal decoupling for ef- ficient burst hdr and restoration

    Tianheng Qiu, Qi Wu, Yuchun Dong, Shenglin Ding, Xuan Huang, Hu Wei, and Guanghua Pan. Recursive multi- exposure alignment with spatiotemporal decoupling for ef- ficient burst hdr and restoration. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 1...

    2025

  32. [40]

    The Second Challenge on Cross-Domain Few-Shot Object Detection at NTIRE 2026: Methods and Results

    Xingyu Qiu, Yuqian Fu, Jiawei Geng, Bin Ren, Jiancheng Pan, Zongwei Wu, Hao Tang, Yanwei Fu, Radu Timo- fte, Nicu Sebe, Mohamed Elhoseiny, et al. The Second Challenge on Cross-Domain Few-Shot Object Detection at NTIRE 2026: Methods and Results . InProceedings of the IEEE/CVF C...

    2026

  33. [41]

    NTIRE 2026 The 3rd Restore Any Image Model (RAIM) Challenge: Multi-Exposure Image Fusion in Dynamic Scenes (Track2)

    Lishen Qu, Yao Liu, Jie Liang, Hui Zeng, Wen Dai, Ya-nan Guan, Guanyi Qin, Shihao Zhou, Jufeng Yang, Lei Zhang, Radu Timofte, et al. NTIRE 2026 The 3rd Restore Any Image Model (RAIM) Challenge: Multi-Exposure Image Fusion in Dynamic Scenes (Track2) . InProceedings of the IEEE/...

    2026

  34. [42]

    Optical flow estima- tion using a spatial pyramid network

    Anurag Ranjan and Michael J Black. Optical flow estima- tion using a spatial pyramid network. InProceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 4161–4170, 2017. 5

    2017

  35. [43]

    The Eleventh NTIRE 2026 Efficient Super-Resolution Challenge Report

    Bin Ren, Hang Guo, Yan Shu, Jiaqi Ma, Ziteng Cui, Shuhong Liu, Guofeng Mei, Lei Sun, Zongwei Wu, Fahad Shahbaz Khan, Salman Khan, Radu Timofte, Yawei Li, et al. The Eleventh NTIRE 2026 Efficient Super-Resolution Challenge Report . InProceedings of the IEEE/CVF Conference on Co...

    2026

  36. [44]

    Conde, Jeffrey Chen, Zhuyun Zhou, Zongwei Wu, Radu Timofte, et al

    Tim Seizinger, Florin-Alexandru Vasluianu, Marcos V . Conde, Jeffrey Chen, Zhuyun Zhou, Zongwei Wu, Radu Timofte, et al. The First Controllable Bokeh Rendering Challenge at NTIRE 2026 . InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) ...

    2026

  37. [45]

    The Third Challenge on Image Denoising at NTIRE 2026: Methods and Results

    Lei Sun, Hang Guo, Bin Ren, Shaolin Su, Xian Wang, Danda Pani Paudel, Luc Van Gool, Radu Timofte, Yawei Li, et al. The Third Challenge on Image Denoising at NTIRE 2026: Methods and Results . InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition ...

    2026

  38. [46]

    The Second Challenge on Event-Based Image Deblurring at NTIRE 2026: Methods and Results

    Lei Sun, Weilun Li, Xian Wang, Zhendong Li, Letian Shi, Dannong Xu, Deheng Zhang, Mengshun Hu, Shuang Guo, Shaolin Su, Radu Timofte, Danda Pani Paudel, Luc Van Gool, et al. The Second Challenge on Event-Based Image Deblurring at NTIRE 2026: Methods and Results . InProceedings ...

    2026

  39. [47]

    NTIRE 2026 The First Challenge on Blind Computational Aberration Correction: Methods and Results

    Lei Sun, Xiaolong Qian, Qi Jiang, Xian Wang, Yao Gao, Kailun Yang, Kaiwei Wang, Radu Timofte, Danda Pani Paudel, Luc Van Gool, et al. NTIRE 2026 The First Challenge on Blind Computational Aberration Correction: Methods and Results . InProceedings of the IEEE/CVF Con- ference o...

    2026

  40. [48]

    In- stance normalization: The missing ingredient for fast styliza- tion.arXiv preprint arXiv:1607.08022, 2016

    Dmitry Ulyanov, Andrea Vedaldi, and Victor Lempitsky. In- stance normalization: The missing ingredient for fast styliza- tion.arXiv preprint arXiv:1607.08022, 2016. 5

    2016 arXiv

  41. [49]

    Learning- Based Ambient Lighting Normalization: NTIRE 2026 Chal- lenge Results and Findings

    Florin-Alexandru Vasluianu, Tim Seizinger, Jeffrey Chen, Zhuyun Zhou, Zongwei Wu, Radu Timofte, et al. Learning- Based Ambient Lighting Normalization: NTIRE 2026 Chal- lenge Results and Findings . InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognit...

    2026

  42. [50]

    Advances in Single- Image Shadow Removal: Results from the NTIRE 2026 Challenge

    Florin-Alexandru Vasluianu, Tim Seizinger, Zhuyun Zhou, Zongwei Wu, Radu Timofte, et al. Advances in Single- Image Shadow Removal: Results from the NTIRE 2026 Challenge . InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Work- shops, 2026. 2

    2026

  43. [51]

    The Second Challenge on Real-World Face Restoration at NTIRE 2026: Methods and Results

    Jingkai Wang, Jue Gong, Zheng Chen, Kai Liu, Jiatong Li, Yulun Zhang, Radu Timofte, et al. The Second Challenge on Real-World Face Restoration at NTIRE 2026: Methods and Results . InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Work- ...

    2026

  44. [52]

    NTIRE 2026 Challenge on 3D Content Super-Resolution: Methods and Results

    Longguang Wang, Yulan Guo, Yingqian Wang, Juncheng Li, Sida Peng, Ye Zhang, Radu Timofte, Minglin Chen, Yi Wang, Qibin Hu, Wenjie Lei, et al. NTIRE 2026 Challenge on 3D Content Super-Resolution: Methods and Results . In Proceedings of the IEEE/CVF Conference on Computer Vi- si...

    2026

  45. [53]

    NTIRE 2026 Challenge on Light Field Image Super-Resolution: Methods and Results

    Yingqian Wang, Zhengyu Liang, Fengyuan Zhang, Wending Zhao, Longguang Wang, Juncheng Li, Jungang Yang, Radu Timofte, Yulan Guo, et al. NTIRE 2026 Challenge on Light Field Image Super-Resolution: Methods and Results . In Proceedings of the IEEE/CVF Conference on Computer Vi- si...

    2026

  46. [54]

    Bovik, Hamid R

    Zhou Wang, Alan C. Bovik, Hamid R. Sheikh, and Eero P. Simoncelli. Image quality assessment: From error visibility to structural similarity.IEEE Transactions on Image Pro- cessing, 13(4):600–612, 2004. 3

    2004

  47. [55]

    Efficient Low Light Image Enhancement: NTIRE 2026 Challenge Report

    Jiebin Yan, Chenyu Tu, Qinghua Lin, Zongwei WU, Weixia Zhang, Zhihua Wang, Peibei Cao, Yuming Fang, Xiaoning Liu, Zhuyun Zhou, Radu Timofte, et al. Efficient Low Light Image Enhancement: NTIRE 2026 Challenge Report . In Proceedings of the IEEE/CVF Conference on Computer Vi- si...

    2026

  48. [56]

    NTIRE 2026 Challenge on High- Resolution Depth of non-Lambertian Surfaces

    Pierluigi Zama Ramirez, Fabio Tosi, Luigi Di Stefano, Radu Timofte, Alex Costanzino, Matteo Poggi, Samuele Salti, Ste- fano Mattoccia, et al. NTIRE 2026 Challenge on High- Resolution Depth of non-Lambertian Surfaces . InProceed- ings of the IEEE/CVF Conference on Computer Visi...

    2026

  49. [57]

    Restormer: Efficient transformer for high-resolution image restoration

    Syed Waqas Zamir, Aditya Arora, Salman Khan, Mu- nawar Hayat, Fahad Shahbaz Khan, and Ming-Hsuan Yang. Restormer: Efficient transformer for high-resolution image restoration. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA, USA, Jun...

    2022

  50. [58]

    Practical blind image denoising via swin- conv-unet and data synthesis.Machine Intelligence Re- search, 20(6):822–836, 2023

    Kai Zhang, Yawei Li, Jingyun Liang, Jiezhang Cao, Yu- lun Zhang, Hao Tang, Deng-Ping Fan, Radu Timofte, and Luc Van Gool. Practical blind image denoising via swin- conv-unet and data synthesis.Machine Intelligence Re- search, 20(6):822–836, 2023. 5

    2023

  51. [59]

    NTIRE 2026 Challenge Report on Anomaly Detection of Face Enhancement for UGC Images

    Yan Zhong, Qiufang Ma, Zhen Wang, Tingting Jiang, Radu Timofte, et al. NTIRE 2026 Challenge Report on Anomaly Detection of Face Enhancement for UGC Images . InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2026. 2

    2026

  52. [60]

    Flow-guided deformable alignment with channel-wise self- attention reconstruct for efficient burst hdr restoration

    Weiyu Zhou, Tao Hu, Yixu Feng, Duwei Dai, Yu Cao, Peng Wu, Wei Dong, Yanning Zhang, and Qingsen Yan. Flow-guided deformable alignment with channel-wise self- attention reconstruct for efficient burst hdr restoration. In Proceedings of the Computer Vision and Pattern Recognitio...

    2025

  53. [61]

    NTIRE 2026 Challenge on Bitstream-Corrupted Video Restoration: Meth- ods and Results

    Wenbin Zou, Tianyi Liu, Kejun Wu, Huiping Zhuang, Zong- wei Wu, Zhuyun Zhou, Radu Timofte, et al. NTIRE 2026 Challenge on Bitstream-Corrupted Video Restoration: Meth- ods and Results . InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR)...

    2026

  54. [62]

    TWILIGHT-COWBOY-CHALLENGE- BASELINE.https://github.com/Zufarovich/ TWILIGHT- COWBOY- CHALLENGE- BASELINE, 2026

    Zufarovich. TWILIGHT-COWBOY-CHALLENGE- BASELINE.https://github.com/Zufarovich/ TWILIGHT- COWBOY- CHALLENGE- BASELINE, 2026. GitHub repository, accessed: 2026-03-14. 5

    2026

Pith tools

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