arxiv: 2604.06623 · v1 · submitted 2026-04-08 · 💻 cs.CV

Recognition: unknown

WeatherRemover: All-in-one Adverse Weather Removal with Multi-scale Feature Map Compression

Weikai Qu , Sijun Liang , Cheng Pan , Zikuan Yang , Guanchi Zhou , Xianjun Fu , Bo Liu , Changmiao Wang

show 1 more author

Ahmed Elazab

Authors on Pith no claims yet

Pith reviewed 2026-05-10 18:38 UTC · model grok-4.3

classification 💻 cs.CV

keywords adverse weather removalimage restorationmulti-weather removalUNetvision transformergating mechanismlightweight modelfeature compression

0 comments

The pith

A lightweight UNet-like model with gating and multi-scale transformers removes rain, snow, and fog from images while using fewer parameters and less memory than prior methods.

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

The paper introduces WeatherRemover as an all-in-one solution for restoring images degraded by multiple weather conditions at once. It builds a UNet-style backbone that incorporates a multi-scale pyramid vision transformer to handle features at varying resolutions and adds gating mechanisms in the feed-forward and downsampling stages to filter out redundant information. Channel-wise attention from convolutional layers sharpens focus on useful details, while linear spatial reduction keeps the cost of attention calculations low. The design targets a practical balance of restoration quality against parameter count, inference speed, and memory footprint so the model can run on devices with limited resources. This matters because most existing multi-weather removal approaches either specialize in one condition or demand heavy computation that limits real-world deployment.

Core claim

The central claim is that a UNet-like network equipped with a multi-scale pyramid vision Transformer, channel-wise attention, and gating mechanisms placed in feed-forward and downsampling phases, together with linear spatial reduction, can selectively suppress redundant features and deliver high-quality removal of rain, snow, and fog effects across diverse conditions while achieving lower parameter size, computational overhead, and memory usage than other all-in-one models.

What carries the argument

Gating mechanisms placed in feed-forward and downsampling phases inside a multi-scale pyramid vision Transformer within a UNet-like structure; these gates selectively address redundancy in feature maps while linear spatial reduction limits attention costs.

If this is right

The model offers a single network that handles rain, snow, and fog removal without needing separate specialized models for each condition.
Resource savings in parameters, computation, and memory make deployment feasible on edge devices and in real-time applications.
Gating enables adaptive selection of essential data during processing, which supports consistent performance across varied weather inputs.
Linear spatial reduction directly lowers the computational load of attention operations without sacrificing the multi-scale feature extraction.

Where Pith is reading between the lines

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

The same feature-compression approach could be tested on other single-image restoration problems such as denoising or low-light enhancement to check if the efficiency gains transfer.
Adding temporal modeling might allow the architecture to process short video clips while keeping weather effects consistent across frames.
If the multi-scale pyramid proves robust, the model could be evaluated on images that combine weather degradation with additional issues like motion blur.
The lightweight footprint suggests direct integration into smartphone camera pipelines for on-device correction before storage.

Load-bearing premise

The gating mechanisms combined with the multi-scale transformer and channel-wise attention will reliably filter redundancy and improve quality on unseen weather conditions without creating artifacts or dropping important image details.

What would settle it

Running the model on a standard multi-weather benchmark dataset and finding either lower PSNR or SSIM scores than existing all-in-one methods or higher parameter count and memory usage would show the claimed efficiency-quality balance does not hold.

Figures

Figures reproduced from arXiv: 2604.06623 by Ahmed Elazab, Bo Liu, Changmiao Wang, Cheng Pan, Guanchi Zhou, Sijun Liang, Weikai Qu, Xianjun Fu, Zikuan Yang.

**Figure 1.** Figure 1: Overview of WeatherRemover structure. The WeatherRemover model is structured around a UNet architecture. The left side of the model is dedicated [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

**Figure 2.** Figure 2: The MSA architecture. Within the MSA, the input matrix passes through a [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Figure 3: The linear SRA flowchart for compressing feature maps. The primary [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

**Figure 5.** Figure 5: Overall structure of MS-PVT. MS-PVT is composed of MSA and [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗

**Figure 6.** Figure 6: Desnowing visual comparisons. From left to right, the sequence of images includes the input images, Restormer [ [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗

**Figure 7.** Figure 7: Real-scene desnowing visual comparisons. The figures from left to right are: Input images, Restormer [ [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗

**Figure 8.** Figure 8: Raindrop removal visual comparisons. From left to right, the images are: Input images, Restormer [ [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗

**Figure 9.** Figure 9: Single-weather deraining and dehazing visual comparisons. From left to right, the images displayed are: The original input images, Restormer [ [PITH_FULL_IMAGE:figures/full_fig_p010_9.png] view at source ↗

**Figure 10.** Figure 10: Real-scene rain and dense fog removal visual comparisons. From left to right: Input image, Restormer [ [PITH_FULL_IMAGE:figures/full_fig_p011_10.png] view at source ↗

**Figure 11.** Figure 11: Model performance vs. resolution. Select raindrops as an example [PITH_FULL_IMAGE:figures/full_fig_p013_11.png] view at source ↗

read the original abstract

Photographs taken in adverse weather conditions often suffer from blurriness, occlusion, and low brightness due to interference from rain, snow, and fog. These issues can significantly hinder the performance of subsequent computer vision tasks, making the removal of weather effects a crucial step in image enhancement. Existing methods primarily target specific weather conditions, with only a few capable of handling multiple weather scenarios. However, mainstream approaches often overlook performance considerations, resulting in large parameter sizes, long inference times, and high memory costs. In this study, we introduce the WeatherRemover model, designed to enhance the restoration of images affected by various weather conditions while balancing performance. Our model adopts a UNet-like structure with a gating mechanism and a multi-scale pyramid vision Transformer. It employs channel-wise attention derived from convolutional neural networks to optimize feature extraction, while linear spatial reduction helps curtail the computational demands of attention. The gating mechanisms, strategically placed within the feed-forward and downsampling phases, refine the processing of information by selectively addressing redundancy and mitigating its influence on learning. This approach facilitates the adaptive selection of essential data, ensuring superior restoration and maximizing efficiency. Additionally, our lightweight model achieves an optimal balance between restoration quality, parameter efficiency, computational overhead, and memory usage, distinguishing it from other multi-weather models, thereby meeting practical application demands effectively. The source code is available at https://github.com/RICKand-MORTY/WeatherRemover.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

WeatherRemover is a gated UNet with multi-scale pyramid vision transformer for all-in-one rain/snow/fog removal whose efficiency claims need ablation tables to stand up.

read the letter

The main takeaway is that this paper introduces WeatherRemover, a UNet-like architecture with a multi-scale pyramid vision transformer, channel-wise attention, and gating in feed-forward and downsampling stages for removing effects of rain, snow, and fog all at once. It positions itself as more efficient than existing multi-weather approaches. The work does a solid job addressing a practical need. Single-condition methods dominate the field, and many all-in-one solutions end up with high parameter counts or slow inference that limit their use in real systems like autonomous vehicles. The choice to add linear spatial reduction for attention and place gates to cut redundancy makes sense on paper for balancing quality and speed. Releasing the source code is a plus for anyone who wants to test or extend it. On the downside, the central assertion of achieving an optimal balance in restoration quality, parameters, compute, and memory lacks the backing it needs. Without ablation experiments that compare the full model against stripped-down versions missing the gating or the pyramid transformer, it's difficult to see how much those additions contribute to selective information processing or reduced overhead. Standard metrics on common datasets would help, but if they're not broken down by component or weather type, the advantage over prior hybrids stays unclear. This kind of paper suits practitioners in computer vision who deal with degraded images in outdoor settings and need deployable solutions rather than theoretical advances. Someone building a pipeline for surveillance or driving assistance might pick up useful design patterns here. I think it deserves peer review. The problem is relevant, the architecture is explained in enough detail to assess, and feedback from referees could strengthen the experimental validation without major changes to the core idea.

Referee Report

2 major / 1 minor

Summary. The paper introduces WeatherRemover, a lightweight UNet-like architecture for all-in-one removal of adverse weather effects (rain, snow, fog) from images. It incorporates gating mechanisms in the feed-forward and downsampling phases, a multi-scale pyramid vision Transformer, channel-wise attention derived from CNNs, and linear spatial reduction to selectively remove redundancy, claiming an optimal balance of restoration quality with low parameter count, FLOPs, inference time, and memory usage that outperforms prior multi-weather models for practical applications. Source code is released.

Significance. If the quantitative claims hold with proper validation, the work could provide a deployable all-in-one solution for real-world image restoration, improving downstream CV tasks under diverse weather without the overhead of specialized models. The open-source code is a clear strength for reproducibility.

major comments (2)

[Model Architecture and Experimental Results] The central claim that the gating mechanisms 'selectively address redundancy' and enable 'superior restoration and maximizing efficiency' (Abstract) is load-bearing but unsupported without ablation studies. No tables compare the full model against variants without gating or without linear spatial reduction on metrics such as PSNR/SSIM across rain/snow/fog datasets or on parameter/FLOP counts.
[Abstract and Results] The assertion of achieving an 'optimal balance' distinguishing it from other multi-weather models lacks any reported quantitative results, baselines, or comparisons (e.g., parameter counts, inference time, memory usage, or restoration metrics) in the abstract or visible experimental support, preventing verification of the efficiency-quality tradeoff.

minor comments (1)

[Abstract] The abstract is lengthy and repetitive; condensing the description of components while retaining the core claims would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We have reviewed the major comments carefully and provide point-by-point responses below. We agree that additional experiments and clarifications will strengthen the paper and will incorporate revisions accordingly.

read point-by-point responses

Referee: [Model Architecture and Experimental Results] The central claim that the gating mechanisms 'selectively address redundancy' and enable 'superior restoration and maximizing efficiency' (Abstract) is load-bearing but unsupported without ablation studies. No tables compare the full model against variants without gating or without linear spatial reduction on metrics such as PSNR/SSIM across rain/snow/fog datasets or on parameter/FLOP counts.

Authors: We acknowledge that explicit ablation studies would provide stronger empirical support for the role of the gating mechanisms in addressing redundancy and for the linear spatial reduction in improving efficiency. The current manuscript describes these components in the architecture and reports overall performance gains, but does not include direct variant comparisons. In the revised manuscript, we will add a new ablation subsection with tables evaluating the full model against versions without gating and without linear spatial reduction. These will include PSNR/SSIM results across the rain, snow, and fog datasets along with parameter counts and FLOPs to quantify the contributions. revision: yes
Referee: [Abstract and Results] The assertion of achieving an 'optimal balance' distinguishing it from other multi-weather models lacks any reported quantitative results, baselines, or comparisons (e.g., parameter counts, inference time, memory usage, or restoration metrics) in the abstract or visible experimental support, preventing verification of the efficiency-quality tradeoff.

Authors: We agree that the abstract would benefit from explicit quantitative support for the efficiency-quality tradeoff claim. While the experimental results section already presents comparisons against prior multi-weather models on restoration metrics (PSNR/SSIM), parameter counts, FLOPs, inference time, and memory usage, these are not summarized in the abstract. In the revision, we will update the abstract to include key numerical highlights of these comparisons, thereby making the 'optimal balance' assertion directly verifiable from the abstract while retaining the detailed tables in the main text. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical architecture proposal with independent experimental validation

full rationale

The paper describes a UNet-like model incorporating gating mechanisms, multi-scale pyramid vision Transformer, channel-wise attention, and linear spatial reduction for multi-weather image restoration. All load-bearing claims concern empirical performance (PSNR, parameter count, FLOPs, memory) after standard training on weather datasets. No equations, predictions, or first-principles results are presented that reduce to fitted inputs by construction. Design choices are motivated by stated goals of redundancy removal and efficiency but are not derived from or equivalent to the target metrics. No self-citation chains, uniqueness theorems, or ansatz smuggling appear in the abstract or described structure. The work is self-contained as an architecture proposal evaluated externally via benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim depends on the unproven effectiveness of the custom gating and attention modules for redundancy reduction; no external benchmarks or proofs are provided in the abstract.

free parameters (1)

model hyperparameters and training settings
Standard neural network weights and optimization choices fitted to weather image datasets.

axioms (1)

domain assumption Gating mechanisms strategically placed within feed-forward and downsampling phases refine information processing by selectively addressing redundancy
Invoked in the abstract as a core design choice without independent justification or ablation evidence.

pith-pipeline@v0.9.0 · 5582 in / 1242 out tokens · 37573 ms · 2026-05-10T18:38:52.928570+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages · 1 internal anchor

[1]

Transweather: Transformer-based restoration of images degraded by adverse weather conditions,

J. M. Jose Valanarasu, R. Yasarla, and V . M. Patel, “Transweather: Transformer-based restoration of images degraded by adverse weather conditions,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2022, pp. 2343–2353

work page 2022
[2]

Focal network for image restoration,

Y . Cui, W. Ren, X. Cao, and A. Knoll, “Focal network for image restoration,” inProc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2023, pp. 12 955–12 965

work page 2023
[3]

Progressive rain removal based on the combination network of cnn and transformer,

T. Wang, K. Wang, and Q. Li, “Progressive rain removal based on the combination network of cnn and transformer,”Comput. Intell. Neurosci., vol. 2022, no. 1, p. 5067175, 2022

work page 2022
[4]

An image is worth 16x16 words: Transformers for image recognition at scale,

A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Transformers for image recognition at scale,” inProc. Int. Conf. Learn. Representations (ICLR), 2021

work page 2021
[5]

Swin transformer: Hierarchical vision transformer using shifted windows,

Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” inProc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2021, pp. 9992–10 002

work page 2021
[6]

Restormer: Efficient transformer for high-resolution image restoration,

S. W. Zamir, A. Arora, S. Khan, M. Hayat, F. S. Khan, and M. Yang, “Restormer: Efficient transformer for high-resolution image restoration,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2022, pp. 5718–5729

work page 2022
[7]

Pyramid vision transformer: A versatile backbone for dense prediction without convolutions,

W. Wang, E. Xie, X. Li, D.-P. Fan, K. Song, D. Liang, T. Lu, P. Luo, and L. Shao, “Pyramid vision transformer: A versatile backbone for dense prediction without convolutions,” inProc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2021, pp. 548–558

work page 2021
[8]

Pvt v2: Improved baselines with pyramid vision transformer,

W. Wang, E. Xie, X. Li, D.-P. Fan, K. Song, D. Liang, T. Lu, P. Luo, and L. Shao, “Pvt v2: Improved baselines with pyramid vision transformer,” Comput. Vis. Media, vol. 8, no. 3, pp. 415–424, 2022

work page 2022
[9]

All in one bad weather removal using architectural search,

R. Li, R. T. Tan, and L.-F. Cheong, “All in one bad weather removal using architectural search,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2020, pp. 3172–3182

work page 2020
[10]

U-net: Convolutional networks for biomedical image segmentation,

O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inProc. Int. Conf. Med. Image Comput. Comput.-Assist. Intervention (MICCAI), vol. 9351, 2015, pp. 234–241

work page 2015
[11]

Learning a sparse transformer network for effective image deraining,

X. Chen, H. Li, M. Li, and J. Pan, “Learning a sparse transformer network for effective image deraining,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2023, pp. 5896–5905

work page 2023
[12]

Attention is all you need,

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. u. Kaiser, and I. Polosukhin, “Attention is all you need,” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 30, 2017

work page 2017
[13]

Single image haze removal using dark channel prior,

K. He, J. Sun, and X. Tang, “Single image haze removal using dark channel prior,”IEEE Trans. Pattern Anal. Mach. Intell. (TPAMI), vol. 33, no. 12, pp. 2341–2353, 2011

work page 2011
[14]

Dual residual networks leveraging the potential of paired operations for image restoration,

X. Liu, M. Suganuma, Z. Sun, and T. Okatani, “Dual residual networks leveraging the potential of paired operations for image restoration,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2019, pp. 7000–7009

work page 2019
[15]

Deep residual learning for image recognition,

K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2016, pp. 770–778

work page 2016
[16]

Lmqformer: A laplace-prior-guided mask query transformer for lightweight snow removal,

J. Lin, N. Jiang, Z. Zhang, W. Chen, and T. Zhao, “Lmqformer: A laplace-prior-guided mask query transformer for lightweight snow removal,”IEEE Trans. Circuits Syst. Video Technol., vol. 33, no. 11, pp. 6225–6235, 2023

work page 2023
[17]

Restoring vision in adverse weather conditions with patch-based denoising diffusion models,

O. ¨Ozdenizci and R. Legenstein, “Restoring vision in adverse weather conditions with patch-based denoising diffusion models,”IEEE Trans. Pattern Anal. Mach. Intell. (TPAMI), vol. 45, no. 8, pp. 10 346–10 357, 2023

work page 2023
[18]

Denoising diffusion probabilistic models,

J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 33, 2020, pp. 6840–6851

work page 2020
[19]

Revitalizing convolutional network for image restoration,

Y . Cui, W. Ren, X. Cao, and A. Knoll, “Revitalizing convolutional network for image restoration,”IEEE Trans. Pattern Anal. Mach. Intell. (TPAMI), pp. 1–16, 2024

work page 2024
[20]

Is space-time attention all you need for video understanding?

G. Bertasius, H. Wang, and L. Torresani, “Is space-time attention all you need for video understanding?” inProc. Int. Conf. Mach. Learn. (ICML),, ser. Proceedings of Machine Learning Research, vol. 139, 2021, pp. 813–824

work page 2021
[21]

Segment anything,

A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo, P. Dolla’r, and R. Girshick, “Segment anything,” inProc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2023, pp. 3992–4003

work page 2023
[22]

End-to-end object detection with transformers,

N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, and S. Zagoruyko, “End-to-end object detection with transformers,” inProc. Eur . Conf. Comput. Vis. (ECCV), 2020, pp. 213–229

work page 2020
[23]

Bs-ldm: Effective bone suppression in high- resolution chest x-ray images with conditional latent diffusion models,

Y . Sun, Z. Chen, H. Zheng, W. Deng, J. Liu, W. Min, A. Elazab, X. Wan, C. Wang, and R. Ge, “Bs-ldm: Effective bone suppression in high- resolution chest x-ray images with conditional latent diffusion models,” 2025,arXiv:2412.15670

work page arXiv 2025
[24]

Gaussian Error Linear Units (GELUs)

D. Hendrycks and K. Gimpel, “Gaussian error linear units (gelus),” 2023, arXiv:1606.08415

work page internal anchor Pith review Pith/arXiv arXiv 2023
[25]

Real-time single image and video super- resolution using an efficient sub-pixel convolutional neural network,

W. Shi, J. Caballero, F. Husz ´ar, J. Totz, A. P. Aitken, R. Bishop, D. Rueckert, and Z. Wang, “Real-time single image and video super- resolution using an efficient sub-pixel convolutional neural network,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2016, pp. 1874–1883

work page 2016
[26]

Improved techniques for training consistency models,

Y . Song and P. Dhariwal, “Improved techniques for training consistency models,” inProc. Int. Conf. Learn. Representations (ICLR), 2024

work page 2024
[27]

Desnownet: Context-aware deep network for snow removal,

Y .-F. Liu, D.-W. Jaw, S.-C. Huang, and J.-N. Hwang, “Desnownet: Context-aware deep network for snow removal,”IEEE Trans. Image Process., vol. 27, no. 6, pp. 3064–3073, 2018

work page 2018
[28]

Attentive generative adversarial network for raindrop removal from a single image,

R. Qian, R. T. Tan, W. Yang, J. Su, and J. Liu, “Attentive generative adversarial network for raindrop removal from a single image,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2018, pp. 2482–2491

work page 2018
[29]

Heavy rain image restoration: Integrating physics model and conditional adversarial learning,

R. Li, L.-F. Cheong, and R. T. Tan, “Heavy rain image restoration: Integrating physics model and conditional adversarial learning,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2019, pp. 1633–1642

work page 2019
[30]

DAWN: Vehicle detection in adverse weather nature dataset,

M. A. Kenk and M. Hassaballah, “Dawn: Vehicle detection in adverse weather nature dataset,” 2020,arXiv:2008.05402

work page arXiv 2020
[31]

Adam: A method for stochastic optimization,

D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” inProc. Int. Conf. Learn. Representations (ICLR), 2015

work page 2015
[32]

Scope of validity of psnr in im- age/video quality assessment,

Q. Huynh-Thu and M. Ghanbari, “Scope of validity of psnr in im- age/video quality assessment,”Electron. Lett., vol. 44, no. 13, pp. 800– 801, 2008

work page 2008
[33]

Image quality assessment: from error visibility to structural similarity,

Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, “Image quality assessment: from error visibility to structural similarity,”IEEE Trans. Image Process., vol. 13, no. 4, pp. 600–612, 2004

work page 2004
[34]

Long short-term memory,

S. Hochreiter and J. Schmidhuber, “Long short-term memory,”Neural Comput., vol. 9, no. 8, pp. 1735–1780, 1997

work page 1997
[35]

Iterative image rain removal network using consecutive residual long short-term memory,

S. Y . Park, T. H. Park, and I. K. Eom, “Iterative image rain removal network using consecutive residual long short-term memory,”Neuro- computing, vol. 589, p. 127752, 2024

work page 2024
[36]

Deep learning for seeing through window with raindrops,

Y . Quan, S. Deng, Y . Chen, and H. Ji, “Deep learning for seeing through window with raindrops,” inProc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2019, pp. 2463–2471

work page 2019
[37]

Image de-raining transformer,

J. Xiao, X. Fu, A. Liu, F. Wu, and Z.-J. Zha, “Image de-raining transformer,”IEEE Trans. Pattern Anal. Mach. Intell. (TPAMI), vol. 45, no. 11, pp. 12 978–12 995, 2023

work page 2023
[38]

Rain-free and residue hand-in-hand: A progressive coupled network for real-time image deraining,

K. Jiang, Z. Wang, P. Yi, C. Chen, Z. Wang, X. Wang, J. Jiang, and C.-W. Lin, “Rain-free and residue hand-in-hand: A progressive coupled network for real-time image deraining,”IEEE Trans. Image Process., vol. 30, pp. 7404–7418, 2021

work page 2021
[39]

Multi-stage progressive image restoration,

S. W. Zamir, A. Arora, S. Khan, M. Hayat, F. S. Khan, M.-H. Yang, and L. Shao, “Multi-stage progressive image restoration,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2021, pp. 14 816–14 826

work page 2021
[40]

Learning without forgetting,

Z. Li and D. Hoiem, “Learning without forgetting,”IEEE Trans. Pattern Anal. Mach. Intell. (TPAMI), vol. 40, no. 12, pp. 2935–2947, 2018

work page 2018
[41]

AIR: Ana- lytic imbalance rectifier for continual learning,

D. Fang, Y . Zhu, Z. Lin, C. Chen, Z. Zeng, and H. Zhuang, “AIR: Ana- lytic imbalance rectifier for continual learning,” 2024,arXiv.2408.10349

work page arXiv 2024