Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Leveraging Content and Context Cues for Low-Light Image Enhancement

T0 review · 4 major / 5 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 Adding CLIP-learned image priors and content/context semantic guidance to a zero-reference enhancer improves machine cognition in low light without paired training data.

desk verdict Useful extension of the authors' CLIP-guided zero-reference enhancement, but the missing train/test split makes the headline gains unsubstantiated. read the letter →

arxiv 2412.07693 v1 pith:KIFODPTJ submitted 2024-12-10 cs.CV eess.IV

Igor Morawski , Kai He , Shusil Dangi , Winston H. Hsu This is my paper · ORCID
classification cs.CVeess.IV
keywords lowlightlow-lightenhancementsemanticguidanceunsupervisedCLIPpromptlearningzero-referencemachinecognition
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

Low-light images confuse computer vision systems, and collecting paired normal-light training data is expensive. This paper claims that a zero-reference enhancer—one that needs no paired or unpaired normal-light data—can be made better for machine cognition by adding two kinds of CLIP-based guidance: a learned image prior and semantic content/context cues. The image prior is learned by contrasting noise-averaged and noise-preserving resampled versions of low-light images, producing prompt embeddings that constrain contrast and noise. The semantic guidance converts existing box annotations into text descriptions of objects inside and outside each training patch, and trains the enhancer to match enhanced patches to those descriptions. Across object detection, face detection, and classification benchmarks, the paper reports consistent task-based gains over related zero-reference methods, along with reduced over-saturation and noise over-amplification.

What carries the argument

The load-bearing mechanism is the joint training loss on a lightweight curve-estimation network, combining (1) the zero-reference losses of Zero-DCE (exposure, spatial consistency, color or Gray-World, and illumination smoothness), (2) a CLIP prompt-pair prior learned by contrasting averaged versus subsampled resamplings of the same low-light images, and (3) two cross-entropy semantic-guidance losses that match enhanced patches, via CLIP image/text embeddings, to abstracted comma-separated descriptions of objects inside the patch (content) and outside the patch (context). The content and context losses are computed on an $N \times N$ similarity matrix in both directions, image-to-text and text-to-image. The prior constrains global contrast and noise amplification; the semantic guidance enriches color and foreground-background separability; together they modify what the curve parameters produce. The method uses Zero-DCE's pixel-wise curve application $L_{En}(x) = L_{En-1}(x) + A_n(x) L_{En-1}(x)(1-L_{En-1}(x))$ and keeps the baseline architecture unchanged.

What would settle it

Train the full method on a strict disjoint split of the same datasets (e.g., train only on explicitly held-out partitions of NOD, ExDark, and DarkFace, and evaluate on the remaining images from those datasets), then compare mAP and classification accuracy with the paper's tables; if the gains over the no-enhancement baseline vanish or shrink sharply, the central claim of generalization fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that injecting a CLIP-learned image prior and content/context semantic guidance into a zero-reference curve-based enhancement network (Zero-DCE) improves the network's output specifically for machine cognition, without any need for paired or unpaired normal-light training data. The learned prior is a pair of text prompts: a positive prompt trained to match m×m-averaged images (a fast proxy for denoising) and a negative prompt trained to match 1:m-subsampled images that retain noise, learned with cross-entropy loss. The semantic guidance is realized as two matching tasks within each training batch: the enhanced image patch must match text descriptions of the object instances inside the patch (content), and also descriptions of instances outside the patch (context). The paper reports that this combination raises mAP and classification accuracy on NOD, NOD SE, LOD, ExDark, and DarkFace relative to the Zero-DCE baseline and other zero-reference methods, while holding inference cost low, and it argues from the results that human-perceptual restoration quality (PSNR/SSIM) does not correlate with downstream task performance in low-light enhancement.

Load-bearing premise

The load-bearing premise is that the images used for training (drawn from NOD, ExDark, DarkFace, ExLPose, LOL, and BAID) are disjoint from the images used for evaluation on NOD, NOD SE, ExDark, and DarkFace; the paper never states a train/test split, so if evaluation images overlap the training set, the reported gains could reflect memorization rather than generalization.

Editorial extensions

If this is right

  • A single enhancement module trained this way can be inserted in front of any downstream detector or classifier, improving low-light performance at negligible inference cost, since the semantic guidance and prompt prior act only at training time.
  • The training recipe requires only low-light images plus existing bounding-box annotations; for paired datasets without annotations, the authors auto-label with an open-vocabulary detector, so the method scales to new categories without fixing the training category set.
  • The same training strategy transfers to other zero-reference baselines (RUAS, SGZ, SCI, Zero-DCE), giving consistent task-based gains on all but SCI, where the authors hypothesize the extra losses interfere with its self-calibration module.
  • With 32× down/up-scaling of the curve parameter map at inference, the method retains most of its task-based gain while cutting computational complexity to 0.08 GMACs, roughly three orders of magnitude below the un-scaled baseline.
  • The reported decoupling of PSNR/SSIM from task accuracy supports the paper's claim that low-light enhancement for machine cognition should be evaluated by downstream task metrics, not by perceptual quality alone.

Reading between the lines

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

  • A testable extension the authors do not run: apply the content/context guidance to other degradations (dehazing, deraining, super-resolution) for machine-vision pipelines; if the semantic matching loss is the active ingredient, it should transfer to any restoration task with box or text annotations.
  • The learned prior's dependence on CLIP as a frozen encoder implies its ceiling is set by CLIP's ability to recognize objects in heavily degraded patches; using a vision-language model trained on low-light or noisy data could push the gains further, which the paper does not explore.
  • The authors' observation that QuadPrior over-smooths extreme low-light instances suggests a complementary direction: combining semantic guidance with an explicit denoising or inpainting module could recover instances that detection models currently miss, rather than relying on the zero-reference curve network which has no denoising capability.
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 / 5 minor

Summary. The paper proposes a zero-reference low-light image enhancement method that builds on Zero-DCE by adding a learned CLIP-based image prior (via prompt learning with an averaging/subsampling augmentation) and two semantic guidance losses based on abstracted content and context descriptions derived from existing bounding-box annotations. The enhancement model is trained without paired or unpaired normal-light data, and the method is evaluated by downstream task performance (object detection, face detection, image classification) on NOD, NOD SE, LOD, ExDark, and DarkFace, as well as by PSNR/SSIM on VE-LOL. The central claim is that the proposed prior and semantic guidance lead to consistent task-based improvements over related zero-reference methods.

Significance. If the claims hold, the work would provide a lightweight, training-time-only enhancement module that improves machine cognition in low light without requiring paired or unpaired normal-light data, while leveraging open-vocabulary CLIP capabilities to avoid fixing the annotated category set. The paper's focus on task-based evaluation with external detectors (YOLO-World, YOLO5Face) and its explicit discussion of the non-correlation between restoration metrics and downstream task performance are strengths, as is the inclusion of an ablation study and a generalization study across multiple zero-reference baselines. However, the central empirical claim is currently compromised by the lack of a stated train/evaluation split and by several internal inconsistencies in the reported results.

major comments (4)
  1. [IV-A and IV-D] The enhancement model is trained on images extracted from NOD, ExDark, DarkFace, ExLPose, LOL, and BAID (Section IV-A), and the CLIP projection layer is fine-tuned on content/context descriptions from these data. The task-based evaluation in Tables III, IV, and VII then reports results on NOD, NOD SE, ExDark, and DarkFace without specifying any train/test split. Because three of the four evaluation benchmarks are part of the training corpus, the reported gains over the no-enhancement baseline and over methods evaluated with their original checkpoints may reflect memorization rather than a general enhancement operator. The authors must state the exact split, release training/evaluation lists, or re-run the comparisons on held-out subsets of these datasets and on datasets not used for training before the central claim can be assessed.
  2. [III-A, Eqs. (1)-(2)] The label assignment in the prompt-learning loss appears inverted relative to the intended semantics. The text states that y = 0 for a positive (averaged) image and y = 1 for a negative (subsampled) image, while Eq. (2) defines ŷ as the softmax probability that an image matches the positive prompt. With the stated assignment, the binary cross-entropy loss pushes the averaged image toward the negative prompt and the subsampled image toward the positive prompt, which is the opposite of the described 'positive quality prompt' and 'negative quality prompt' behavior. Please clarify whether this is a typographical error in the text or an actual property of the implementation; if the implementation follows the equations as written, the learned prior would be the inverse of what is claimed and the qualitative benefits attributed to it would be unexplained.
  3. [IV-D, Table VII] The comparison against paired and unpaired methods is inequitable because the competitors (SNR+SKF, DRBN+SKF, PairLIE, DiffLL, CLIP-LIT, PIE, QuadPrior) are evaluated using their original public checkpoints, whereas Ours is trained on the evaluation datasets themselves (NOD, ExDark, DarkFace, and others). The paper's statement that a cross-dataset setup was also reported in recent works does not remedy the asymmetry when the proposed method is trained on the target benchmarks. This setup conflates the effect of the proposed enhancement losses with the advantage of in-distribution training, so the rankings in Table VII cannot be interpreted as evidence of the method's generalization.
  4. [IV-B and IV-D, Tables IV and IX] The claim of 'consistent improvements' over the baseline is not supported by the reported numbers. In Table IV, Ours achieves 42.5% classification accuracy on ExDark, below the Zero-DCE baseline's 44.1%; in Table VII, PairLIE reaches 47.7% on the same metric, while Ours obtains 42.5%. Moreover, Table IX shows that applying the proposed strategy to SCI decreases mAP on NOD (43.9% to 43.7%), NOD SE (27.0% to 26.7%), and LOD (45.6% to 45.4%), which contradicts the sentence in Section IV-G that states 'we observe significant improvement in all methods with exception of SCI.' The authors should revise the consistency claim and explicitly discuss these exceptions.
minor comments (5)
  1. [III-B, Eq. (5)] The exposure loss is written with an undefined double sum '1/M \sum_{i=1}^M \sum_{i=1}^X |\hat{I}_i - E|'; it should presumably be a single sum over the M non-overlapping 16x16 patches, with the index consistent throughout the equation.
  2. [III-B, Eq. (10)] Equation (10) has a missing closing parenthesis in 'cos(Φ_img(Î_i), Φ_txt(A_j)' and should read 'cos(Φ_img(Î_i), Φ_txt(A_j))'.
  3. [Fig. 3 caption] The phrase 'with any limitations on annotated object categories' appears to be a typo and should read 'without any limitations on annotated object categories'.
  4. [III-B, first paragraph] The sentence 'Focusing on leveraging the CLIP [17] model for semantic segmentation' is confusing, since the paper does not perform semantic segmentation; please rephrase to describe the actual use of CLIP for semantic guidance.
  5. [IV-D and Tables III-VII] The abbreviation 'NOD SE' is used in the tables and text but is never defined; please define it in Section IV (it appears to refer to a subset or extreme-light variant of NOD).

Circularity Check

0 steps flagged · score 2.0 of 10

No construction-level circularity: the central task-based claims are empirical and evaluated with external detectors; only minor, non-load-bearing self-citations are present.

full rationale

The paper's central claim is that adding a learned CLIP image prior plus content/context semantic guidance to Zero-DCE improves downstream task performance. This claim is empirical and is tested against external detectors (YOLO-World, YOLO5Face) and external baselines. The derivation chain is: (1) learn positive/negative prompts by classifying averaged versus subsampled low-light patches; (2) train the enhancement network with Zero-DCE's zero-reference losses plus the learned prompt prior and content/context cross-entropy classification losses; (3) evaluate on object detection, face detection, and classification benchmarks. No step defines its output in terms of its input: the learned prompt is a regularizer trained on the same low-light distribution, not a fitted value later reported as a prediction; the content/context losses are training objectives, not evaluation metrics; and the reported improvements are measured on external downstream tasks, so they do not reduce to the loss definitions by construction. The paper does cite the authors' prior work [16] for the earlier semantic-guidance idea and uses the authors' NOD dataset [24] as a benchmark, but these self-citations are not load-bearing for the claimed improvement, which is supported by comparisons against external methods and by external detectors. A separate validity concern exists: the paper never states a train/test split, and three of the evaluation datasets (NOD, ExDark, DarkFace) overlap the announced training corpus, so some reported gains could reflect in-sample evaluation. That is a data-leakage/correctness risk, not a construction-level circularity. Accordingly, no circular step is identified; the score of 2 reflects only the presence of minor, non-load-bearing self-citation.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

No new physical entities are introduced. The central claim rests on CLIP as a semantic teacher, on Zero-DCE natural-image assumptions, and on an unstated disjoint-data assumption that is the main fragility. Free parameters are hand-set hyperparameters inherited from the baseline plus the prompt-augmentation scheme; none are fitted to the downstream task metrics beyond training the enhancement network itself.

free parameters (5)
  • Exposure target E = 0.6
    Hand-set expected brightness in the zero-reference exposure loss, Eq. (5), inherited from Zero-DCE [13].
  • Zero-reference loss weights = lambda_exp=10, lambda_spa=1, lambda_RGB=5, lambda_TV=200
    Set manually in Section IV-A; these control the balance between exposure, spatial consistency, color, and illumination smoothness.
  • Prompt length N = 16
    Chosen in Section IV-A for the learned positive and negative CLIP prompt pair.
  • Sampling scale factor s = 4
    Used in Section III-A and IV-A: 4x4 averaging for positive samples and 1:4 subsampling for negative samples.
  • Automatic annotation confidence threshold = 30%
    Used in Section IV-A when generating pseudo-labels for paired datasets without manual annotation.
assumptions (6)
  • domain assumption CLIP text-image cosine similarity is a valid semantic signal for guiding low-light enhancement
    The content/context losses and the learned prior loss in Sections III-B1 and III-B2 make gradients depend on CLIP matching enhanced patches to text descriptions; if CLIP is insensitive to these cues in noisy low-light images, the guidance is noise.
  • ad hoc to paper Average pooling acts as a denoising proxy for a positive image and 1:4 subsampling preserves noise for a negative image
    Section III-A and Figure 5 define the prompt-learning targets; this mapping from resampling to image quality is chosen by the authors and is not independently validated.
  • standard math Zero-reference losses based on Gray-World, exposure, and smoothness assumptions are valid for machine-cognition-oriented enhancement
    The baseline Zero-DCE losses in Section III-B are taken unchanged from [13]; the paper inherits their assumptions about natural images.
  • domain assumption Bounding-box annotations converted to comma-separated object lists capture content and context, and itemization order does not affect CLIP matching
    Section III-B2 and the pilot study assert that order has minimal effect; the content and context guidance depends on this.
  • domain assumption Training and evaluation datasets are disjoint
    Section IV-A builds a training set from NOD, ExDark, DarkFace, ExLPose, LOL and BAID, while Section IV-D evaluates on NOD, NOD SE, ExDark and DarkFace. No split is stated; the comparative tables assume no overlap.
  • domain assumption Detection and classification accuracy of YOLO-World and YOLO5Face adequately measures machine-cognition benefit
    Section IV-A selects these detectors and the paper interprets their mAP and accuracy as the target objective for enhancement.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Leveraging Content and Context Cues for Low-Light Image Enhancement." pith.science (2026). https://pith.science/paper/KIFODPTJ

@misc{pith2026241207693,
  author       = {Pith},
  title        = {Pith review of: Leveraging Content and Context Cues for Low-Light Image Enhancement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KIFODPTJ}},
  note         = {Machine review of arXiv:2412.07693}
}
read the original abstract

Low-light conditions have an adverse impact on machine cognition, limiting the performance of computer vision systems in real life. Since low-light data is limited and difficult to annotate, we focus on image processing to enhance low-light images and improve the performance of any downstream task model, instead of fine-tuning each of the models which can be prohibitively expensive. We propose to improve the existing zero-reference low-light enhancement by leveraging the CLIP model to capture image prior and for semantic guidance. Specifically, we propose a data augmentation strategy to learn an image prior via prompt learning, based on image sampling, to learn the image prior without any need for paired or unpaired normal-light data. Next, we propose a semantic guidance strategy that maximally takes advantage of existing low-light annotation by introducing both content and context cues about the image training patches. We experimentally show, in a qualitative study, that the proposed prior and semantic guidance help to improve the overall image contrast and hue, as well as improve background-foreground discrimination, resulting in reduced over-saturation and noise over-amplification, common in related zero-reference methods. As we target machine cognition, rather than rely on assuming the correlation between human perception and downstream task performance, we conduct and present an ablation study and comparison with related zero-reference methods in terms of task-based performance across many low-light datasets, including image classification, object and face detection, showing the effectiveness of our proposed method.

Figures

Figures reproduced from arXiv: 2412.07693 by the authors.

Figure 1
Figure 1. As we target machine cognition, rather than rely on assuming the [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Timeline including recent developments in task-based optimization of low-light image enhancement methods, datasets dedicated to low-light image [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 2
Figure 2. We also summarize advantages of sRGB and raw sensor [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figures from the paper (10 more)
Figure 3
Figure 3. Figure 3: Our proposed method. In the first stage, we propose to learn the positive and negative image priors without any need for paired or unpaired normal-light [PITH_FULL_IMAGE:figures/full_fig_p005_3.png]
Figure 4
Figure 4. Figure 4: Since our motivation is to make the maximal use of the existing low-light annotation, which may be difficult or costly to obtain, we apply two separate [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: We use image sampling to augment positive and negative image prompt [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 7
Figure 7. Figure 7: Average image pixel intensities in our training dataset, by sample [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: Sources of samples in our training dataset. [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 9
Figure 9. Figure 9: Ablation study of our proposed method. The learned prompt improves the overall image contrast, helps reducing under- and overexposure and reduces [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: Qualitative comparison with related methods. [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]
Figure 11
Figure 11. Figure 11: Impact of low light on object detection performance, measured in [PITH_FULL_IMAGE:figures/full_fig_p011_11.png]
Figure 12
Figure 12. Figure 12: Qualitative results of object detection using Yolo-World model [77] on the NOD [24] and LOD [25]. Instead of assuming correlation between human [PITH_FULL_IMAGE:figures/full_fig_p012_12.png]
Figure 13
Figure 13. Figure 13: Prompted by task-based evaluation results in Tab. VII, we additionally [PITH_FULL_IMAGE:figures/full_fig_p013_13.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. CURVE: CLIP-Utilized Reinforcement Learning for Visual Image Enhancement via Simple Image Processing

    cs.CV 2025-05 conditional novelty 4.0 of 10

    An RL agent iteratively adjusts a Bezier tone curve, rewarded by CLIP text-image similarity, for fast zero-reference low-light and multi-exposure image enhancement.

Reference graph

Works this paper leans on

82 extracted references · 53 canonical work pages · cited by 1 Pith paper

  1. [16]

    Unsupervised image prior via prompt learning and clip semantic guidance for low-light image enhancement,

    I. Morawski, K. He, S. Dangi, and W. H. Hsu, “Unsupervised image prior via prompt learning and clip semantic guidance for low-light image enhancement,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 5971–5981

    work page 2024

  2. [1]

    Getting to know low-light images with the exclusively dark dataset,

    Y . P. Loh and C. S. Chan, “Getting to know low-light images with the exclusively dark dataset,” Computer Vision and Image Understanding , vol. 178, pp. 30–42, 2019

    2019

  3. [2]

    Learning to restore low- light images via decomposition-and-enhancement,

    K. Xu, X. Yang, B. Yin, and R. W. Lau, “Learning to restore low- light images via decomposition-and-enhancement,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 2281–2290. 14

    2020

  4. [3]

    Adaptive unfolding total variation network for low-light image enhancement,

    C. Zheng, D. Shi, and W. Shi, “Adaptive unfolding total variation network for low-light image enhancement,” in Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 4439– 4448

    work page 2021

  5. [4]

    Half wavelet attention on m- net+ for low-light image enhancement,

    C.-M. Fan, T.-J. Liu, and K.-H. Liu, “Half wavelet attention on m- net+ for low-light image enhancement,” in 2022 IEEE International Conference on Image Processing (ICIP) . IEEE, 2022, pp. 3878–3882

    work page 2022

  6. [5]

    Snr-aware low-light image enhancement,

    X. Xu, R. Wang, C.-W. Fu, and J. Jia, “Snr-aware low-light image enhancement,” in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 17 693–17 703

    work page 2022

  7. [6]

    Learning semantic-aware knowledge guidance for low-light image en- hancement,

    Y . Wu, C. Pan, G. Wang, Y . Yang, J. Wei, C. Li, and H. T. Shen, “Learning semantic-aware knowledge guidance for low-light image en- hancement,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , June 2023, pp. 1662–1671

    work page 2023

  8. [7]

    Low-light image enhancement via structure modeling and guidance,

    X. Xu, R. Wang, and J. Lu, “Low-light image enhancement via structure modeling and guidance,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , June 2023, pp. 9893–9903

    work page 2023

Show all 82 references
  1. [8]

    Deep retinex decomposition for low-light enhancement,

    C. Wei, W. Wang, W. Yang, and J. Liu, “Deep retinex decomposition for low-light enhancement,” arXiv preprint arXiv:1808.04560 , 2018

    2018 arXiv

  2. [9]

    Kindling the darkness: A practical low-light image enhancer,

    Y . Zhang, J. Zhang, and X. Guo, “Kindling the darkness: A practical low-light image enhancer,” inProceedings of the 27th ACM international conference on multimedia , 2019, pp. 1632–1640

    2019

  3. [10]

    Beyond brightening low-light images,

    Y . Zhang, X. Guo, J. Ma, W. Liu, and J. Zhang, “Beyond brightening low-light images,” International Journal of Computer Vision , vol. 129, pp. 1013–1037, 2021

    2021

  4. [11]

    Sparse gradient reg- ularized deep retinex network for robust low-light image enhancement,

    W. Yang, W. Wang, H. Huang, S. Wang, and J. Liu, “Sparse gradient reg- ularized deep retinex network for robust low-light image enhancement,” IEEE Transactions on Image Processing, vol. 30, pp. 2072–2086, 2021

    2021

  5. [12]

    Deep color consistent network for low-light image enhancement,

    Z. Zhang, H. Zheng, R. Hong, M. Xu, S. Yan, and M. Wang, “Deep color consistent network for low-light image enhancement,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2022, pp. 1899–1908

    2022

  6. [13]

    Zero- reference deep curve estimation for low-light image enhancement,

    C. Guo, C. Li, J. Guo, C. C. Loy, J. Hou, S. Kwong, and R. Cong, “Zero- reference deep curve estimation for low-light image enhancement,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 1780–1789

    2020

  7. [14]

    Learning to enhance low-light image via zero-reference deep curve estimation,

    C. Li, C. Guo, and C. C. Loy, “Learning to enhance low-light image via zero-reference deep curve estimation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 8, pp. 4225–4238, 2021

    2021

  8. [15]

    Semantic-guided zero-shot learning for low- light image/video enhancement,

    S. Zheng and G. Gupta, “Semantic-guided zero-shot learning for low- light image/video enhancement,” in Proceedings of the IEEE/CVF Winter conference on applications of computer vision , 2022, pp. 581– 590

    2022

  9. [17]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark et al., “Learning transferable visual models from natural language supervision,” in International conference on machine learning . PMLR, 2021, pp. 8748–8763

    2021

  10. [18]

    Learning to see in the dark,

    C. Chen, Q. Chen, J. Xu, and V . Koltun, “Learning to see in the dark,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 3291–3300

    2018

  11. [19]

    What else can fool deep learning? ad- dressing color constancy errors on deep neural network performance,

    M. Afifi and M. S. Brown, “What else can fool deep learning? ad- dressing color constancy errors on deep neural network performance,” in Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 243–252

    2019

  12. [20]

    Advancing image understanding in poor visibility environments: A collective benchmark study,

    W. Yang, Y . Yuan, W. Ren, J. Liu, W. J. Scheirer, Z. Wang, Zhang, and et al., “Advancing image understanding in poor visibility environments: A collective benchmark study,”IEEE Transactions on Image Processing, vol. 29, pp. 5737–5752, 2020

    2020

  13. [21]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in neural information processing systems , vol. 33, pp. 6840– 6851, 2020

    2020

  14. [22]

    Hardware-in-the-loop end-to-end optimization of camera im- age processing pipelines,

    A. Mosleh, A. Sharma, E. Onzon, F. Mannan, N. Robidoux, and F. Heide, “Hardware-in-the-loop end-to-end optimization of camera im- age processing pipelines,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2020, pp. 7529–7538

    2020

  15. [23]

    Recurrent exposure generation for low-light face detection,

    J. Liang, J. Wang, Y . Quan, T. Chen, J. Liu, H. Ling, and Y . Xu, “Recurrent exposure generation for low-light face detection,” IEEE Transactions on Multimedia , vol. 24, pp. 1609–1621, 2021

    2021

  16. [24]

    Nod: Taking a closer look at detection under extreme low-light conditions with night object detection dataset,

    I. Morawski, Y .-A. Chen, Y .-S. Lin, and W. H. Hsu, “Nod: Taking a closer look at detection under extreme low-light conditions with night object detection dataset,” arXiv preprint arXiv:2110.10364 , 2021

    2021 arXiv

  17. [25]

    Crafting object detection in very low light,

    Y . Hong, K. Wei, L. Chen, and Y . Fu, “Crafting object detection in very low light,” in BMVC, vol. 1, 2021, p. 3

    2021

  18. [26]

    Genisp: neural isp for low-light machine cognition,

    I. Morawski, Y .-A. Chen, Y .-S. Lin, S. Dangi, K. He, and W. H. Hsu, “Genisp: neural isp for low-light machine cognition,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 630–639

    2022

  19. [27]

    Exploring clip for assessing the look and feel of images,

    J. Wang, K. C. Chan, and C. C. Loy, “Exploring clip for assessing the look and feel of images,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 37, 2023, pp. 2555–2563

    2023

  20. [28]

    Iterative prompt learning for unsupervised backlit image enhancement,

    Z. Liang, C. Li, S. Zhou, R. Feng, and C. C. Loy, “Iterative prompt learning for unsupervised backlit image enhancement,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 8094–8103

    2023

  21. [29]

    Tienet: task-oriented image enhancement network for degraded object detection,

    Y . Wang, J. Guo, R. Wang, W. He, and C. Li, “Tienet: task-oriented image enhancement network for degraded object detection,” Signal, Image and Video Processing , pp. 1–8, 2023

    2023

  22. [30]

    Featen- hancer: Enhancing hierarchical features for object detection and beyond under low-light vision,

    K. A. Hashmi, G. Kallempudi, D. Stricker, and M. Z. Afzal, “Featen- hancer: Enhancing hierarchical features for object detection and beyond under low-light vision,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 6725–6735

    2023

  23. [31]

    Raw or cooked? object detection on raw images,

    W. Ljungbergh, J. Johnander, C. Petersson, and M. Felsberg, “Raw or cooked? object detection on raw images,” in Scandinavian Conference on Image Analysis . Springer, 2023, pp. 374–385

    2023

  24. [32]

    Dynamicisp: dy- namically controlled image signal processor for image recognition,

    M. Yoshimura, J. Otsuka, A. Irie, and T. Ohashi, “Dynamicisp: dy- namically controlled image signal processor for image recognition,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 12 866–12 876

    2023

  25. [33]

    Human pose estimation in extremely low-light condi- tions,

    S. Lee, J. Rim, B. Jeong, G. Kim, B. Woo, H. Lee, and S. K. Sunghyun Cho, “Human pose estimation in extremely low-light condi- tions,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2023

    2023

  26. [34]

    Darkvision: a benchmark for low-light image/video perception,

    B. Zhang, Y . Guo, R. Yang, Z. Zhang, J. Xie, J. Suo, and Q. Dai, “Darkvision: a benchmark for low-light image/video perception,” arXiv preprint arXiv:2301.06269, 2023

    2023 arXiv

  27. [35]

    Diff-retinex: Rethinking low-light image enhancement with a generative diffusion model,

    X. Yi, H. Xu, H. Zhang, L. Tang, and J. Ma, “Diff-retinex: Rethinking low-light image enhancement with a generative diffusion model,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 12 302–12 311

    2023

  28. [36]

    Zero-reference low-light enhance- ment via physixcal quadruple priors,

    W. Wang, H. Yang, J. Fu, and J. Liu, “Zero-reference low-light enhance- ment via physixcal quadruple priors,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 26 057–26 066

    2024

  29. [37]

    The retinex theory of color vision,

    E. H. Land, “The retinex theory of color vision,” Scientific american , vol. 237, no. 6, pp. 108–129, 1977

    1977

  30. [38]

    Properties and performance of a center/surround retinex,

    D. J. Jobson, Z.-u. Rahman, and G. A. Woodell, “Properties and performance of a center/surround retinex,” IEEE transactions on image processing, vol. 6, no. 3, pp. 451–462, 1997

    1997

  31. [39]

    A multiscale retinex for bridging the gap between color images and the human observation of scenes,

    ——, “A multiscale retinex for bridging the gap between color images and the human observation of scenes,” IEEE Transactions on Image processing, vol. 6, no. 7, pp. 965–976, 1997

    1997

  32. [40]

    Naturalness preserved enhancement algorithm for non-uniform illumination images,

    S. Wang, J. Zheng, H.-M. Hu, and B. Li, “Naturalness preserved enhancement algorithm for non-uniform illumination images,” IEEE transactions on image processing , vol. 22, no. 9, pp. 3538–3548, 2013

    2013

  33. [41]

    A fusion-based enhancing method for weakly illuminated images,

    X. Fu, D. Zeng, Y . Huang, Y . Liao, X. Ding, and J. Paisley, “A fusion-based enhancing method for weakly illuminated images,” Signal Processing, vol. 129, pp. 82–96, 2016

    2016

  34. [42]

    Lime: Low-light image enhancement via illumination map estimation,

    X. Guo, Y . Li, and H. Ling, “Lime: Low-light image enhancement via illumination map estimation,” IEEE Transactions on image processing , vol. 26, no. 2, pp. 982–993, 2016

    2016

  35. [43]

    Joint denoising and enhancement for low-light images via retinex model,

    M. Li, J. Liu, W. Yang, and Z. Guo, “Joint denoising and enhancement for low-light images via retinex model,” in International Forum on Digital TV and Wireless Multimedia Communications . Springer, 2017, pp. 91–99

    2017

  36. [44]

    Structure-revealing low- light image enhancement via robust retinex model,

    M. Li, J. Liu, W. Yang, X. Sun, and Z. Guo, “Structure-revealing low- light image enhancement via robust retinex model,” IEEE Transactions on Image Processing , vol. 27, no. 6, pp. 2828–2841, 2018

    2018

  37. [45]

    Low- light image enhancement based on virtual exposure,

    W. Wang, D. Yan, X. Wu, W. He, Z. Chen, X. Yuan, and L. Li, “Low- light image enhancement based on virtual exposure,” Signal Processing: Image Communication, vol. 118, p. 117016, 2023

    2023

  38. [46]

    Enlightengan: Deep light enhancement without paired supervision,

    Y . Jiang, X. Gong, D. Liu, Y . Cheng, C. Fang, X. Shen, J. Yang, P. Zhou, and Z. Wang, “Enlightengan: Deep light enhancement without paired supervision,” IEEE transactions on image processing, vol. 30, pp. 2340– 2349, 2021

    2021

  39. [47]

    Abandoning the bayer-filter to see in the dark,

    X. Dong, W. Xu, Z. Miao, L. Ma, C. Zhang, J. Yang, Z. Jin, A. B. J. Teoh, and J. Shen, “Abandoning the bayer-filter to see in the dark,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 17 431–17 440. 15

    2022

  40. [48]

    Retinex-inspired unrolling with cooperative prior architecture search for low-light image enhance- ment,

    R. Liu, L. Ma, J. Zhang, X. Fan, and Z. Luo, “Retinex-inspired unrolling with cooperative prior architecture search for low-light image enhance- ment,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 10 561–10 570

    2021

  41. [49]

    Low-light image enhancement with normalizing flow,

    Y . Wang, R. Wan, W. Yang, H. Li, L.-P. Chau, and A. Kot, “Low-light image enhancement with normalizing flow,” in Proceedings of the AAAI conference on artificial intelligence , vol. 36, 2022, pp. 2604–2612

    2022

  42. [50]

    Uretinex- net: Retinex-based deep unfolding network for low-light image enhance- ment,

    W. Wu, J. Weng, P. Zhang, X. Wang, W. Yang, and J. Jiang, “Uretinex- net: Retinex-based deep unfolding network for low-light image enhance- ment,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2022, pp. 5901–5910

    2022

  43. [51]

    Toward fast, flexible, and robust low-light image enhancement,

    L. Ma, T. Ma, R. Liu, X. Fan, and Z. Luo, “Toward fast, flexible, and robust low-light image enhancement,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 5637–5646

    2022

  44. [52]

    You do not need additional priors or regularizers in retinex-based low-light image enhancement,

    H. Fu, W. Zheng, X. Meng, X. Wang, C. Wang, and H. Ma, “You do not need additional priors or regularizers in retinex-based low-light image enhancement,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 18 125–18 134

    2023

  45. [53]

    Learning a deep single image contrast enhancer from multi-exposure images,

    J. Cai, S. Gu, and L. Zhang, “Learning a deep single image contrast enhancer from multi-exposure images,” IEEE Transactions on Image Processing, vol. 27, no. 4, pp. 2049–2062, 2018

    2018

  46. [54]

    Seeing motion in the dark,

    C. Chen, Q. Chen, M. N. Do, and V . Koltun, “Seeing motion in the dark,” in Proceedings of the IEEE/CVF International conference on computer vision, 2019, pp. 3185–3194

    2019

  47. [55]

    R2rnet: Low-light image enhancement via real-low to real-normal network,

    J. Hai, Z. Xuan, R. Yang, Y . Hao, F. Zou, F. Lin, and S. Han, “R2rnet: Low-light image enhancement via real-low to real-normal network,” Journal of Visual Communication and Image Representation , vol. 90, p. 103712, 2023

    2023

  48. [56]

    Learning photo- graphic global tonal adjustment with a database of input/output image pairs,

    V . Bychkovsky, S. Paris, E. Chan, and F. Durand, “Learning photo- graphic global tonal adjustment with a database of input/output image pairs,” in CVPR 2011. IEEE, 2011, pp. 97–104

    2011

  49. [57]

    Learning to see moving objects in the dark,

    H. Jiang and Y . Zheng, “Learning to see moving objects in the dark,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019, pp. 7324–7333

    2019

  50. [58]

    From fidelity to perceptual quality: A semi-supervised approach for low-light image enhancement,

    W. Yang, S. Wang, Y . Fang, Y . Wang, and J. Liu, “From fidelity to perceptual quality: A semi-supervised approach for low-light image enhancement,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 3063–3072

    2020

  51. [59]

    Pie: Physics-inspired low-light enhancement,

    D. Liang, Z. Xu, L. Li, M. Wei, and S. Chen, “Pie: Physics-inspired low-light enhancement,” International Journal of Computer Vision , pp. 1–22, 2024

    2024

  52. [60]

    Learning a simple low-light image enhancer from paired low-light instances,

    Z. Fu, Y . Yang, X. Tu, Y . Huang, X. Ding, and K.-K. Ma, “Learning a simple low-light image enhancer from paired low-light instances,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 22 252–22 261

    2023

  53. [61]

    A spatial processor model for object colour perception,

    G. Buchsbaum, “A spatial processor model for object colour perception,” Journal of the Franklin institute , vol. 310, no. 1, pp. 1–26, 1980

    1980

  54. [62]

    Difflight: Integrating content and detail for low-light image enhancement,

    Y . Feng, S. Hou, H. Lin, Y . Zhu, P. Wu, W. Dong, J. Sun, Q. Yan, and Y . Zhang, “Difflight: Integrating content and detail for low-light image enhancement,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops , June 2024, pp....

    2024

  55. [63]

    Low-light image enhancement with wavelet-based diffusion models,

    H. Jiang, A. Luo, H. Fan, S. Han, and S. Liu, “Low-light image enhancement with wavelet-based diffusion models,” ACM Transactions on Graphics (TOG) , vol. 42, no. 6, pp. 1–14, 2023

    2023

  56. [64]

    Pyramid diffusion models for low-light image enhancement,

    D. Zhou, Z. Yang, and Y . Yang, “Pyramid diffusion models for low-light image enhancement,” arXiv preprint arXiv:2305.10028 , 2023

    2023 arXiv

  57. [65]

    Low-light image enhancement with knowledge distillation,

    Z. Li, Y . Wang, and J. Zhang, “Low-light image enhancement with knowledge distillation,” Neurocomputing, vol. 518, pp. 332–343, 2023

    2023

  58. [66]

    Hla-face: Joint high-low adaptation for low light face detection,

    W. Wang, W. Yang, and J. Liu, “Hla-face: Joint high-low adaptation for low light face detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 16 195–16 204

    2021

  59. [67]

    Arid: A new dataset for recognizing action in the dark,

    Y . Xu, J. Yang, H. Cao, K. Mao, J. Yin, and S. See, “Arid: A new dataset for recognizing action in the dark,” in Deep Learning for Human Activity Recognition: Second International Workshop, DL-HAR 2020, Held in Conjunction with IJCAI-PRICAI 2020, Kyoto, Japan, January 8, 2021...

    2020

  60. [68]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 770–778

    2016

  61. [69]

    When image denoising meets high-level vision tasks: A deep learning approach,

    D. Liu, B. Wen, X. Liu, Z. Wang, and T. S. Huang, “When image denoising meets high-level vision tasks: A deep learning approach,” arXiv preprint arXiv:1706.04284 , 2017

    2017 arXiv

  62. [70]

    Se- mantic segmentation guided real-world super-resolution,

    A. Aakerberg, A. S. Johansen, K. Nasrollahi, and T. B. Moeslund, “Se- mantic segmentation guided real-world super-resolution,” inProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2022, pp. 449–458

    2022

  63. [71]

    Recovering realistic texture in image super-resolution by deep spatial feature transform,

    X. Wang, K. Yu, C. Dong, and C. C. Loy, “Recovering realistic texture in image super-resolution by deep spatial feature transform,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 606–615

    2018

  64. [72]

    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,” in European conference on computer vision . Springer, 2020, pp. 399–415

    2020

  65. [73]

    End-to-end high dynamic range camera pipeline optimiza- tion,

    N. Robidoux, L. E. G. Capel, D.-e. Seo, A. Sharma, F. Ariza, and F. Heide, “End-to-end high dynamic range camera pipeline optimiza- tion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 6297–6307

    2021

  66. [74]

    Open-vocabulary detr with conditional matching,

    Y . Zang, W. Li, K. Zhou, C. Huang, and C. C. Loy, “Open-vocabulary detr with conditional matching,” in European Conference on Computer Vision. Springer, 2022, pp. 106–122

    2022

  67. [75]

    F-vlm: Open-vocabulary object detection upon frozen vision and language models,

    W. Kuo, Y . Cui, X. Gu, A. Piergiovanni, and A. Angelova, “F-vlm: Open-vocabulary object detection upon frozen vision and language models,” arXiv preprint arXiv:2209.15639 , 2022

    2022 arXiv

  68. [76]

    Extract free dense labels from clip,

    C. Zhou, C. C. Loy, and B. Dai, “Extract free dense labels from clip,” in European Conference on Computer Vision. Springer, 2022, pp. 696– 712

    2022

  69. [77]

    Yolo- world: Real-time open-vocabulary object detection,

    T. Cheng, L. Song, Y . Ge, W. Liu, X. Wang, and Y . Shan, “Yolo- world: Real-time open-vocabulary object detection,” arXiv preprint arXiv:2401.17270, 2024

    2024 arXiv

  70. [78]

    Objects365: A large-scale, high-quality dataset for object detection,

    S. Shao, Z. Li, T. Zhang, C. Peng, G. Yu, X. Zhang, J. Li, and J. Sun, “Objects365: A large-scale, high-quality dataset for object detection,” in Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 8430–8439

    2019

  71. [79]

    Backlitnet: A dataset and network for backlit image enhancement,

    X. Lv, S. Zhang, Q. Liu, H. Xie, B. Zhong, and H. Zhou, “Backlitnet: A dataset and network for backlit image enhancement,” Computer Vision and Image Understanding , vol. 218, p. 103403, 2022

    2022

  72. [80]

    Yolo5face: Why reinventing a face detector,

    D. Qi, W. Tan, Q. Yao, and J. Liu, “Yolo5face: Why reinventing a face detector,” 2021

    2021

  73. [81]

    Benchmarking low-light image enhancement and beyond,

    J. Liu, D. Xu, W. Yang, M. Fan, and H. Huang, “Benchmarking low-light image enhancement and beyond,” International Journal of Computer Vision, vol. 129, pp. 1153–1184, 2021

    2021

  74. [82]

    Neighbor2neighbor: Self- supervised denoising from single noisy images,

    T. Huang, S. Li, X. Jia, H. Lu, and J. Liu, “Neighbor2neighbor: Self- supervised denoising from single noisy images,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 14 781–14 790. 16 Igor Morawski is currently working toward his...

    2021

Pith tools

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