Pith. sign in

REVIEW 6 major objections 6 minor 41 references

DeepSPG: Exploring Deep Semantic Prior Guidance for Low-light Image Enhancement with Multimodal Learning

T0 review · 6 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

Pith's one-line read DeepSPG claims that steering Retinex reflectance refinement with image-level segmentation features and text-level CLIP alignment improves low-light enhancement, reporting state-of-the-art PSNR/SSIM on four of five benchmarks with 0.55M…

desk verdict A solid LLIE engineering contribution whose real claims—semantic guidance helps, compact trainable model—are undercut by an overstated SOTA claim and an efficiency comparison that omits the frozen HRNet backbone. read the letter →

arxiv 2504.19127 v1 pith:CSJFKTCJ submitted 2025-04-27 cs.CV cs.MM

classification cs.CVcs.MM
keywords low-lightimageenhancementRetinexdecompositionsemanticpriorguidancemultimodallearningCLIPalignmentsegmentationfeaturesreflectancerefinementrestoration
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

DeepSPG sets out to show that a low-light enhancement network recovers more detail when it is told what the scene contains, not just how dark it is. The paper builds on Retinex decomposition, separating each image into illumination (brightness) and reflectance (object colors and textures independent of lighting), $X_{in}=X_l\cdot X_r$, and refines the reflectance branch under two semantic priors: hierarchical features from a frozen semantic segmentation network (HRNet-W48), and text-level constraints from CLIP, a pretrained vision-language model, that pull the enhanced output toward the description "high-light image" and away from "low-light image." In the reported experiments, the large variant records the highest PSNR on four of the five tested datasets (LOL-v1, LOL-v2-real, LOL-v2-synthetic, SID), while the base model runs at 0.55M parameters, and the ablations attribute gains of about 1.7 dB and 0.022 SSIM to adding the two semantic priors plus coarse-to-fine training. If the claim holds, it means high-level semantic knowledge is not decoration but a practical way to stabilize color and structure in the darkest parts of an image.

What carries the argument

The machinery is Retinex decomposition plus two aligned semantic streams. The governing identity is $X_{in}=X_l\cdot X_r$: the illumination map carries brightness and is adjusted, while the reflectance map carries object color and structure and is refined under semantic guidance. The fusion point is the semantic-aware embedding module (SEM), which projects reflectance and segmentation features into key/query/value spaces, computes the softmax correlation $A^b=\mathrm{Softmax}(W_k(F_i^b)\times W_q(F_s^b)/\sqrt{C})$, and adds a residual feed-forward refinement. The text stream adds a cosine-similarity loss $L_{mul}=D_{cos}(X_{out},T_L)-D_{cos}(X_{out},T_H)$ that separates the enhanced image from the "low-light" prompt and toward the "high-light" prompt, and the full objective is a weighted sum of pixel, edge, semantic-KL, and multimodal losses over a coarse-to-fine multi-scale structure.

What would settle it

Take the darkest, noisiest quartile of SID test images, run the frozen segmentation model on the raw low-light inputs, and compare its output against ground-truth segmentation. If PSNR stays high even where the segmentation map is wrong, while removing the semantic module barely changes performance, then the image-level prior is not load-bearing; the paper's own reported failure case is the natural starting point.

Watch

Extended reading notes

Core claim

The central claim is that semantic priors can be made load-bearing in low-light enhancement by attaching them to the reflectance map rather than to the final image alone. After Retinex decomposition, DeepSPG feeds the reflectance branch with multi-scale features and a segmentation map from a frozen HRNet-W48 segmentation model, using a semantic-aware embedding module whose attention correlation reweights reflectance features by their similarity to semantic prototypes. A CLIP-based text loss then pushes the enhanced image's embedding toward the "high-light image" prompt and away from "low-light image," while a KL-divergence semantic loss keeps the enhanced image's segmentation map aligned with the ground truth's. On the paper's numbers, this combination gives DeepSPG-large 27.03 dB PSNR on LOL-v1 and 28.08 dB on LOL-v2-synthetic, surpassing the previous best by 1.87 dB and 2.41 dB respectively, and the base model reports 26.87 dB and 27.11 dB on the same two datasets with only 0.55M parameters.

Load-bearing premise

The load-bearing premise is that a segmentation network trained on ordinary well-lit images can still read what objects are in an extremely dark, noisy photo; if it cannot, both the image-level guidance and the semantic loss are steering the enhancement with wrong information.

Editorial extensions

If this is right

  • In the reported ablations, adding the image-level semantic prior raises PSNR by 1.06 dB and SSIM by 0.013, and adding the text-level prior adds another 0.34 dB and 0.003, so each modality contributes measurably to the final result.
  • The large variant's reported 27.03 dB on LOL-v1 and 28.08 dB on LOL-v2-synthetic exceed the strongest prior baseline by 1.87 dB and 2.41 dB, indicating that the semantic-prior design closes a real gap on paired benchmarks.
  • With 0.55M parameters and 6.91 G FLOPs, the base model is lighter than the transformer baselines it is compared with, so semantic guidance is compatible with efficient deployment.
  • Because both semantic priors come from frozen pretrained models, the framework inherits their limitations: the paper states that when the segmentation model misreads noisy dark input, enhancement fails to recover correct semantics and colors.

Reading between the lines

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

  • A testable extension the paper leaves implicit is that swapping in a noise-robust segmentation model should transfer directly to gains in extreme dark scenes, since the paper's own failure case is attributed to wrong segmentation features under noise.
  • The text prior is a single global contrast between "low-light image" and "high-light image"; per-scene prompts describing the dominant objects could give the vision-language loss finer control than the current pair.
  • The semantic loss compares segmentation maps, so the benefit may be concentrated in scenes with recognizable objects such as people, text, and indoor objects; on textureless or uniformly dark regions, pixel and edge losses may carry most of the weight, a split the paper does not measure.
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

6 major / 6 minor

Summary. This paper proposes DeepSPG, a low-light image enhancement framework built on Retinex decomposition that incorporates image-level semantic priors from a frozen HRNet-W48 segmentation model and text-level priors from CLIP. The framework uses a semantic-aware embedding module (SEM) to fuse hierarchical semantic features into the reflectance branch and a multimodal loss to align enhanced images with text prompts. Experiments are reported on five benchmark datasets (LOL-v1, LOL-v2-real, LOL-v2-syn, SID, SMID) with quantitative comparisons against eleven methods, ablations, and visual results. The authors claim state-of-the-art performance and better computational efficiency.

Significance. If the results are taken at face value, DeepSPG provides a useful demonstration that high-level semantic priors can improve low-light enhancement, especially in dark, information-depleted regions. The paper ships public code, which supports reproducibility, and the ablation study provides evidence that both image-level and text-level priors individually contribute. However, the strength of the empirical claim is weakened by several discrepancies between the reported numbers and the statements made in the text, and by the omission of the frozen backbone from the efficiency accounting. The core idea is interesting and the experimental structure is sound; the issues are mostly in the reporting and verification, though they are load-bearing for the SOTA and efficiency claims.

major comments (6)
  1. [Sec. 4.2, Table 1] The caption says 'Our DeepSPG significantly outperforms SOTA algorithms,' but the table does not support this for the base model: DeepSPG trails Retinexformer on LOL-v2-real (22.78 vs 22.80 PSNR) and SMID (28.15 vs 29.15 PSNR), and ties on SID (24.44 vs 24.44). DeepSPG-large also trails Retinexformer on SMID (29.08 vs 29.15 PSNR). The sentence 'achieving PSNR improvement in three datasets (LOL-v1, LOL-v2-syn, and SID)' is therefore incorrect for SID. Please correct these claims and either reposition the contribution as competitive rather than SOTA, or conduct additional experiments to resolve the discrepancies.
  2. [Sec. 3.3, Eq. (1); Sec. 4.2, Table 1] The complexity numbers in Table 1 (6.91 GFLOPs, 0.55M params for DeepSPG; 26.03 GFLOPs, 2.13M for DeepSPG-large) count only the trainable subnetwork. At inference the image-level prior requires a forward pass of the frozen HRNet-W48 segmentation model (Eq. 1), which has tens of millions of parameters and substantial FLOPs. Without counting this backbone, the claim of a 'better trade-off between performance and computational costs' is unsupported and is not an apples-to-apples comparison with baselines that do not use such a backbone. Please report end-to-end inference cost (params, FLOPs, and latency) for the full model, or clearly and consistently state that the reported numbers are for the trainable part only, and adjust the efficiency claim accordingly.
  3. [Sec. 4.3, Table 2] The coarse-to-fine (C2F) training scheme, which the ablation credits with a 0.28 dB PSNR improvement, is never defined anywhere in the paper. It appears only in Table 2 and in the caption of Fig. 2. Without a description of the scheme (e.g., whether it refers to progressive resolution training, curriculum over stages, or something else), the ablation is not reproducible and the contribution of this component cannot be assessed. Please add a precise definition and implementation details.
  4. [Sec. 3.5, Eq. (13)] The semantic loss is not clearly specified. Eq. (1) defines X_seg as the segmentation map of the low-light input, while the text in Sec. 3.2 says the loss is computed 'between the generated and ground-truth segmentation maps.' Eq. (13) writes Lsem = KL(X_seg || \hat X_seg) without defining \hat X_seg. If X_seg is the input's segmentation, the asymmetric KL would pull the enhanced image's semantic distribution toward the low-light input's distribution, which may be wrong in exactly the noise-disturbed cases the paper acknowledges in Sec. 5. Please define \hat X_seg explicitly (e.g., segmentation of the enhanced image or of the ground truth), state the direction of the KL divergence used in practice, and justify why this choice is beneficial rather than harmful.
  5. [Sec. 2.2, Sec. 4.2] SKF [14] is the most closely related semantic-guided LLIE method and provides the SEM module that DeepSPG adopts. Yet SKF is not included in the quantitative comparison in Table 1. Since the paper claims to surpass previous methods and specifically to improve on 'image-level semantics without multimodal learning,' the omission of SKF leaves the advantage over the direct predecessor undemonstrated. Please add SKF to the benchmarks or explain why it cannot be compared.
  6. [Sec. 4.1, Table 1] The architecture of DeepSPG-large is never described. Table 1 reports results for DeepSPG-large and the text uses them to support the SOTA claim, but the paper does not state how DeepSPG-large differs from the base model (e.g., width, depth, or input resolution). Without this information, the large-variant results are not reproducible. Please add a description of the variant or remove it from the main comparison.
minor comments (6)
  1. [Fig. 4 caption] The caption of Fig. 4 says 'our PiCat' where it should say 'our DeepSPG.' Please correct this typo, which appears to be a leftover from another manuscript.
  2. [Sec. 4.2] The efficiency comparison mixes metrics: '2.3 times lighter' refers to FLOPs (6.91 vs 15.57 GFLOPs), while '113 times more efficient' refers to FLOPs of MIRNet, but the parameter counts give different ratios (0.55/1.61 is about 2.9x). Please state which quantity is meant and report both params and FLOPs consistently.
  3. [Sec. 3.5] The values of the loss weights λ1, λ2, λ3, and λ4 are not reported, even though they are said to be 'empirically tuned.' Please include the exact values used in the experiments.
  4. [Sec. 1] The name 'Yann LeCun' is misspelled as 'Yan LeCun' in the quotation at the start of the introduction.
  5. [Throughout] The paper contains leftover ACM template artifacts, including 'Conference acronym 'XX,' the ACM copyright block, and '2018' in the reference format. These should be removed or updated for a clean submission.
  6. [Sec. 3.3] The terms 'reflection' and 'reflectance' are used interchangeably (e.g., 'reflection maps' vs 'reflectance map'). Please standardize the terminology.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: DeepSPG's enhancement gains are measured against external PSNR/SSIM benchmarks and supported by component ablations; self-citations are not load-bearing.

full rationale

DeepSPG's central claim is an empirical state-of-the-art result on LOL-v1, LOL-v2, SID, and SMID, reported in Table 1 as PSNR/SSIM against ground-truth images. The image-level semantic prior (Eqs. 1, 4, 7-8) uses a frozen HRNet-W48 segmentation network, and the text-level prior (Eqs. 2-3, 9-10, 14) uses a frozen CLIP model; neither of these inputs encodes the target PSNR or SSIM values, so the reported metrics are not defined in terms of the priors. The semantic loss (Eq. 13) does share the same HRNet backbone that produces the image-level prior, creating an internal coupling between the feature extractor and one training objective, but it does not force the external enhancement metrics: the ablation in Table 2 shows that removing image semantics costs 1.06 dB PSNR and removing text semantics costs another 0.34 dB, which is consistent with the components contributing independently rather than being renamed versions of the evaluation. The self-citations, e.g., [21], [22], [24], [30], and [41], appear in related-work context or as future-work suggestions and are not used to justify the SOTA claim. The stated limitation in Sec. 5, that the pretrained segmentation model can fail under heavy noise, is an acknowledged assumption about the external semantic extractor, not an import of the enhancement result. The omission of frozen HRNet-W48 from the FLOPs/params comparison in Table 1 is an efficiency-reporting concern rather than circularity. Overall, no equation or prediction reduces by construction to its own inputs, so the derivation is self-contained against external benchmarks.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard supervised learning assumptions plus two pre-trained models whose reliability on dark inputs is assumed. No new physical or conceptual entity is introduced. The main free parameters are the unreported loss weights and the hand-chosen CLIP prompts, along with the undefined C2F training scheme.

free parameters (2)
  • Loss weights lambda1, lambda2, lambda3, lambda4
    Eq. 15 defines the total loss as a weighted sum; Sec. 4.1 says the weights are 'empirically tuned' but their values are never reported, so all benchmark numbers depend on unreported choices.
  • CLIP text prompts = low-light image; high-light image
    Sec. 4.1 fixes the prompts to 'low-light image' and 'high-light image'; the paper does not ablate alternative prompts, so the multimodal loss depends on this hand-chosen pair.
assumptions (4)
  • domain assumption Retinex decomposition X_in = X_l * X_r holds for the input images (Eq. 6).
    The entire pipeline assumes each low-light image factorizes into illumination and reflectance layers, which is a modeling approximation of real image formation.
  • domain assumption The frozen HRNet-W48 segmentation model provides useful semantic features on dark, noisy low-light inputs.
    Eq. 1 and Sec. 3.3 apply a segmentation model trained on ADE20K directly to low-light images; the authors' own Sec. 5 says this fails in some extremely noisy scenes.
  • domain assumption CLIP text-image similarity meaningfully distinguishes 'high-light image' from 'low-light image' for enhancement purposes.
    Sec. 3.4 and Eq. 14 rely on CLIP embeddings to push the enhanced image toward a high-light text description; there is no validation that this objective correlates with perceptual or metric quality beyond the reported experiments.
  • ad hoc to paper The coarse-to-fine (C2F) training scheme is a well-defined procedure that improves results by 0.28 dB PSNR (Tab. 2).
    C2F is listed as a component and ablated as Variant 3, but no description of the scheme appears in Sec. 3 or Sec. 4, making it an unstated design choice that affects the central claim.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DeepSPG: Exploring Deep Semantic Prior Guidance for Low-light Image Enhancement with Multimodal Learning." pith.science (2026). https://pith.science/paper/CSJFKTCJ

@misc{pith2026250419127,
  author       = {Pith},
  title        = {Pith review of: DeepSPG: Exploring Deep Semantic Prior Guidance for Low-light Image Enhancement with Multimodal Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CSJFKTCJ}},
  note         = {Machine review of arXiv:2504.19127}
}
read the original abstract

There has long been a belief that high-level semantics learning can benefit various downstream computer vision tasks. However, in the low-light image enhancement (LLIE) community, existing methods learn a brutal mapping between low-light and normal-light domains without considering the semantic information of different regions, especially in those extremely dark regions that suffer from severe information loss. To address this issue, we propose a new deep semantic prior-guided framework (DeepSPG) based on Retinex image decomposition for LLIE to explore informative semantic knowledge via a pre-trained semantic segmentation model and multimodal learning. Notably, we incorporate both image-level semantic prior and text-level semantic prior and thus formulate a multimodal learning framework with combinatorial deep semantic prior guidance for LLIE. Specifically, we incorporate semantic knowledge to guide the enhancement process via three designs: an image-level semantic prior guidance by leveraging hierarchical semantic features from a pre-trained semantic segmentation model; a text-level semantic prior guidance by integrating natural language semantic constraints via a pre-trained vision-language model; a multi-scale semantic-aware structure that facilitates effective semantic feature incorporation. Eventually, our proposed DeepSPG demonstrates superior performance compared to state-of-the-art methods across five benchmark datasets. The implementation details and code are publicly available at https://github.com/Wenyuzhy/DeepSPG.

Figures

Figures reproduced from arXiv: 2504.19127 by the authors.

Figure 1
Figure 1. Comparisons between our DeepSPG (bottom) and [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overall framework of DeepSPG. (a) The main coarse-to-fine enhancement pipeline incorporates both image-level [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Visual results on LOL-v1 [15] (top) and SID [33] (bottom). Brightness correction is equally applied to all cropped patches (blue box and pink box) for better detail comparison. Previous methods often fail due to noise, color distortion, or producing blurry and under- or over-exposed images. In contrast, our DeepSPG effectively removes noise and reconstructs well-exposed image details. (Please Zoom in for the best vi… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Visual comparisons of Restormer [6], MIRNet [37], SNR-Net [5], Retinexformer [4], HAIR [39], and our DeepSPG on some extremely low-light and noisy scenes of the SID [33] dataset. Previous methods often fail due to noise, color distortion, or producing blurry and under-…
Figure 6
Figure 6. Figure 6: Failure case of our DeepSPG on the SID dataset [ [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 5
Figure 5. Figure 5: Visual results of break-down ablations on the LOL [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 35 canonical work pages

  1. [14]

    Learning semantic-aware knowledge guidance for low-light image enhancement

    Yuhui Wu, Chen Pan, Guoqing Wang, Yang Yang, Jiwei Wei, Chongyi Li, and Heng Tao Shen. Learning semantic-aware knowledge guidance for low-light image enhancement. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 1662–1671, 2023

  2. [1]

    Comparing deep learning models for low-light natural scene image enhancement and their impact on object detection and classification: Overview, empirical evaluation, and challenges

    Rayan Al Sobbahi and Joe Tekli. Comparing deep learning models for low-light natural scene image enhancement and their impact on object detection and classification: Overview, empirical evaluation, and challenges. Signal Processing: Image Communication, 109:116848, 2022

  3. [2]

    Degrade for upgrade: Learning degradation representations for real-world low-light image enhancement

    Lintao Xu, Changhui Hu, Weihong Zhu, Fei Wu, Ziyun Cai, Mengjun Ye, and Xiaobo Lu. Degrade for upgrade: Learning degradation representations for real-world low-light image enhancement. Computers and Electrical Engineering , 119:109622, 2024

  4. [3]

    Improving low-light image recognition performance based on image-adaptive learnable module

    Seitaro Ono, Yuka Ogino, Takahiro Toizumi, Atsushi Ito, and Masato Tsukada. Improving low-light image recognition performance based on image-adaptive learnable module. arXiv preprint arXiv:2401.06438, 2024

  5. [4]

    Retinexformer: One-stage retinex-based transformer for low-light image en- hancement

    Yuanhao Cai, Hao Bian, Jing Lin, Haoqian Wang, Radu Timofte, and Yulun Zhang. Retinexformer: One-stage retinex-based transformer for low-light image en- hancement. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 12504–12513, 2023

  6. [5]

    Snr-aware low-light image enhancement

    Xiaogang Xu, Ruixing Wang, Chi-Wing Fu, and Jiaya Jia. Snr-aware low-light image enhancement. In Proceedings of the IEEE/CVF International conference on computer vision, 2022

  7. [6]

    Restormer: Efficient transformer for high- resolution image restoration

    Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shah- baz Khan, and Ming-Hsuan Yang. Restormer: Efficient transformer for high- resolution image restoration. In Proceedings of the IEEE/CVF International confer- ence on computer vision , 2022

  8. [7]

    Band representation-based semi-supervised low-light image enhancement: Bridging the gap between signal fidelity and perceptual quality

    Wenhan Yang, Shiqi Wang, Yuming Fang, Yue Wang, and Jiaying Liu. Band representation-based semi-supervised low-light image enhancement: Bridging the gap between signal fidelity and perceptual quality. IEEE Transactions on Image Processing, 2021

Show all 41 references
  1. [8]

    A dynamic histogram equalization for image contrast enhancement

    Mohammad Abdullah-Al-Wadud, Md Hasanul Kabir, M Ali Akber Dewan, and Oksam Chae. A dynamic histogram equalization for image contrast enhancement. IEEE transactions on consumer electronics , 53(2):593–600, 2007

  2. [9]

    Local color correction using non-linear masking

    Nathan Moroney. Local color correction using non-linear masking. In Color and Imaging conference, volume 8, pages 108–111. Society of Imaging Science and Technology, 2000

  3. [10]

    Image enhancement by modified contrast-stretching manip- ulation

    Ching-Chung Yang. Image enhancement by modified contrast-stretching manip- ulation. Optics & Laser Technology, 38(3):196–201, 2006

  4. [11]

    Llcnn: A convolutional neural network for low-light image enhancement

    Li Tao, Chuang Zhu, Guoqing Xiang, Yuan Li, Huizhu Jia, and Xiaodong Xie. Llcnn: A convolutional neural network for low-light image enhancement. In 2017 IEEE Visual Communications and Image Processing (VCIP) , pages 1–4. IEEE, 2017

  5. [12]

    Rscnn: A cnn-based method to enhance low-light remote-sensing images

    Linshu Hu, Mengjiao Qin, Feng Zhang, Zhenhong Du, and Renyi Liu. Rscnn: A cnn-based method to enhance low-light remote-sensing images. Remote Sensing, 13(1):62, 2020

  6. [13]

    Tpet: two-stage perceptual enhancement transformer network for low-light image enhancement.Engineering Applications of Artificial Intelligence, 116:105411, 2022

    Hengshuai Cui, Jinjiang Li, Zhen Hua, and Linwei Fan. Tpet: two-stage perceptual enhancement transformer network for low-light image enhancement.Engineering Applications of Artificial Intelligence, 116:105411, 2022

  7. [15]

    The retinex theory of color vision

    Edwin H Land. The retinex theory of color vision. Scientific american, 1977

  8. [16]

    Lime: Low-light image enhancement via illumination map estimation

    Xiaojie Guo, Yu Li, and Haibin Ling. Lime: Low-light image enhancement via illumination map estimation. IEEE Transactions on image processing , 26(2):982– 993, 2016

  9. [17]

    Naturalness preserved image enhancement using a priori multi-layer lightness statistics

    Shuhang Wang and Gang Luo. Naturalness preserved image enhancement using a priori multi-layer lightness statistics. IEEE transactions on image processing , 27(2):938–948, 2017

  10. [18]

    Llnet: A deep autoencoder approach to natural low-light image enhancement

    Kin Gwn Lore, Adedotun Akintayo, and Soumik Sarkar. Llnet: A deep autoencoder approach to natural low-light image enhancement. Pattern Recognition, 61:650– 662, 2017

  11. [19]

    Deep photo enhancer: Unpaired learning for image enhancement from photographs with gans

    Yu-Sheng Chen, Yu-Ching Wang, Man-Hsin Kao, and Yung-Yu Chuang. Deep photo enhancer: Unpaired learning for image enhancement from photographs with gans. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 6306–6314, 2018

  12. [20]

    Beyond brightening low-light images

    Yonghua Zhang, Xiaojie Guo, Jiayi Ma, Wei Liu, and Jiawan Zhang. Beyond brightening low-light images. International Journal of Computer Vision, 129:1013– 1037, 2021

  13. [21]

    Learning physics-informed color-aware transforms for low-light image enhancement

    Xingxing Yang, Jie Chen, and Zaifeng Yang. Learning physics-informed color-aware transforms for low-light image enhancement. arXiv preprint arXiv:2504.11896, 2025

  14. [22]

    Multi-scale progressive feature embedding for accurate nir-to-rgb spectral domain translation

    Xingxing Yang, Jie Chen, and Zaifeng Yang. Multi-scale progressive feature embedding for accurate nir-to-rgb spectral domain translation. In 2023 IEEE International Conference on Visual Communications and Image Processing (VCIP) , pages 1–5, 2023

  15. [23]

    Multimodal low-light image enhancement with depth information

    Zhen Wang, Dongyuan Li, Guang Li, Ziqing Zhang, and Renhe Jiang. Multimodal low-light image enhancement with depth information. In Proceedings of the 32nd ACM International Conference on Multimedia , pages 4976–4985, 2024

  16. [24]

    Cooperative colorization: Exploring latent cross-domain priors for nir image spectrum translation

    Xingxing Yang, Jie Chen, and Zaifeng Yang. Cooperative colorization: Exploring latent cross-domain priors for nir image spectrum translation. In Proceedings of the 31st ACM International Conference on Multimedia , pages 2409–2417, 2023

  17. [25]

    Enlightengan: Deep light enhancement without paired supervision

    Yifan Jiang, Xinyu Gong, Ding Liu, Yu Cheng, Chen Fang, Xiaohui Shen, Jianchao Yang, Pan Zhou, and Zhangyang Wang. Enlightengan: Deep light enhancement without paired supervision. IEEE transactions on image processing , 30:2340–2349, 2021

  18. [26]

    Implicit neu- ral representation for cooperative low-light image enhancement

    Shuzhou Yang, Moxuan Ding, Yanmin Wu, Zihan Li, and Jian Zhang. Implicit neu- ral representation for cooperative low-light image enhancement. In Proceedings of the IEEE/CVF international conference on computer vision , pages 12918–12927, 2023

  19. [27]

    Unsupervised image prior via prompt learning and clip semantic guidance for low-light image en- hancement

    Igor Morawski, Kai He, Shusil Dangi, and Winston H Hsu. Unsupervised image prior via prompt learning and clip semantic guidance for low-light image en- hancement. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5971–5981, 2024

  20. [28]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning , p...

  21. [29]

    Deep high- resolution representation learning for visual recognition

    Jingdong Wang, Ke Sun, Tianheng Cheng, Borui Jiang, Chaorui Deng, Yang Zhao, Dong Liu, Yadong Mu, Mingkui Tan, Xinggang Wang, et al. Deep high- resolution representation learning for visual recognition. IEEE transactions on pattern analysis and machine intelligence , 43(10):33...

  22. [30]

    Hyperspectral image reconstruction via combinatorial embedding of cross-channel spatio-spectral clues

    Xingxing Yang, Jie Chen, and Zaifeng Yang. Hyperspectral image reconstruction via combinatorial embedding of cross-channel spatio-spectral clues. In Proceed- ings of the AAAI Conference on Artificial Intelligence , volume 38, pages 6567–6575, 2024

  23. [31]

    Deep retinex decom- position for low-light enhancement

    Chen Wei, Wenjing Wang, Wenhan Yang, and Jiaying Liu. Deep retinex decom- position for low-light enhancement. In Proceedings of the British Machine Vision Conference, 2018

  24. [32]

    Sparse gradient regularized deep retinex network for robust low-light image enhancement

    Wenhan Yang, Wenjing Wang, Haofeng Huang, Shiqi Wang, and Jiaying Liu. Sparse gradient regularized deep retinex network for robust low-light image enhancement. IEEE Transactions on Image Processing , 2021

  25. [33]

    Learning to see in the dark

    Chen Chen, Qifeng Chen, Jia Xu, and Vladlen Koltun. Learning to see in the dark. In Proceedings of the IEEE/CVF International conference on computer vision , 2018

  26. [34]

    Seeing motion in the dark

    Chen Chen, Qifeng Chen, Minh N Do, and Vladlen Koltun. Seeing motion in the dark. In Proceedings of the IEEE/CVF International conference on computer vision , pages 3185–3194, 2019

  27. [35]

    Unpaired image enhancement featuring reinforcement-learning-controlled image editing software

    Satoshi Kosugi and Toshihiko Yamasaki. Unpaired image enhancement featuring reinforcement-learning-controlled image editing software. In AAAI, 2020

  28. [36]

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

    Ke Xu, Xin Yang, Baocai Yin, and Rynson WH Lau. Learning to restore low-light images via decomposition-and-enhancement. In Proceedings of the IEEE/CVF International conference on computer vision , 2020

  29. [37]

    Learning enriched features for real image restoration and enhancement

    Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, Ming-Hsuan Yang, and Ling Shao. Learning enriched features for real image restoration and enhancement. In European Conferenceon Computer Vision , 2020

  30. [38]

    Ultra-high-definition low-light image enhancement: A benchmark and transformer-based method

    Tao Wang, Kaihao Zhang, Tianrun Shen, Wenhan Luo, Bjorn Stenger, and Tong Lu. Ultra-high-definition low-light image enhancement: A benchmark and transformer-based method. In Association for the Advancement of Artificial Intelligence, volume 37, pages 2654–2662, 2023

  31. [39]

    Hair: Hypernetworks- based all-in-one image restoration

    Jin Cao, Yi Cao, Li Pang, Deyu Meng, and Xiangyong Cao. Hair: Hypernetworks- based all-in-one image restoration. arXiv preprint arXiv:2408.08091, 2024

  32. [40]

    Segment anything

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alexander C Berg, Wan-Yen Lo, et al. Segment anything. In Proceedings of the IEEE/CVF international conference on computer vision, pages 4015–4026, 2023

  33. [41]

    Colormamba: Towards high-quality nir-to-rgb spectral translation with mamba

    Huiyu Zhai, Guang Jin, Xingxing Yang, and Guosheng Kang. Colormamba: Towards high-quality nir-to-rgb spectral translation with mamba. arXiv preprint arXiv:2408.08087, 2024

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.