pith. machine review for the scientific record. sign in

arxiv: 2604.05687 · v2 · submitted 2026-04-07 · 💻 cs.CV

Recognition: 2 theorem links

· Lean Theorem

3D Smoke Scene Reconstruction Guided by Vision Priors from Multimodal Large Language Models

Xinye Zheng , Fei Wang , Yiqi Nie , Kun Li , Junjie Chen , Jiaqi Zhao , Yanyan Wei , Zhiliang Wu
Authors on Pith no claims yet

Pith reviewed 2026-05-10 19:40 UTC · model grok-4.3

classification 💻 cs.CV
keywords 3D Gaussian Splattingsmoke scene reconstructionnovel view synthesisimage enhancementmultimodal large language modelsview-dependent medium modelingparticipating media
0
0 comments X

The pith

Multimodal model priors plus a view-dependent medium branch let 3D Gaussian splatting reconstruct smoke scenes and synthesize clear novel views.

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

The paper develops a method to reconstruct 3D scenes from multi-view images that have been heavily degraded by smoke, which scatters light and destroys cross-view consistency. It first applies an MLLM called Nano-Banana-Pro to enhance the degraded images and supply clearer observations. It then introduces Smoke-GS, an extension of 3D Gaussian Splatting that adds a lightweight view-dependent medium branch to model how smoke appearance changes with viewing direction. A sympathetic reader would care because standard reconstruction pipelines break down in smoky conditions that arise in fires, pollution monitoring, and visual effects, and this approach aims to restore usable geometry and novel views while keeping the fast rendering speed of Gaussian splatting.

Core claim

The authors claim that vision priors from multimodal large language models can be used to enhance smoke-degraded images, after which a medium-aware 3D Gaussian Splatting framework called Smoke-GS can reconstruct the underlying scene and produce restoration-oriented novel views. Smoke-GS represents the scene with explicit 3D Gaussians and adds a lightweight view-dependent medium branch that captures direction-dependent scattering variations caused by smoke, all without requiring explicit physical medium models or ground-truth medium data.

What carries the argument

Smoke-GS, a 3D Gaussian Splatting model augmented with a lightweight view-dependent medium branch that captures and compensates for direction-dependent smoke scattering effects using only MLLM-enhanced images.

If this is right

  • Consistent and visually clear novel views can be synthesized from smoke-degraded multi-view inputs.
  • Rendering speed remains comparable to standard 3D Gaussian Splatting while handling smoke degradation.
  • Reconstruction succeeds without additional ground-truth medium observations or full physical modeling of scattering.
  • Cross-view consistency lost to view-dependent smoke effects is restored through the combination of enhanced images and the medium branch.

Where Pith is reading between the lines

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

  • The same lightweight medium branch could be adapted to other participating media such as fog, dust, or underwater scattering by retraining only the branch.
  • Because the approach relies on off-the-shelf MLLM priors rather than task-specific training data, it may generalize to additional degradation types with little extra supervision.
  • The efficiency preserved by the Gaussian splatting backbone suggests the method could support real-time reconstruction in dynamic smoky environments if the medium branch stays lightweight.

Load-bearing premise

The lightweight view-dependent medium branch can accurately model smoke scattering effects from enhanced images alone, without ground-truth medium data or explicit physical simulation.

What would settle it

Quantitative comparison of novel-view PSNR, SSIM, and perceptual metrics on real smoke scenes against ground-truth clear captures; if removing the medium branch or the MLLM enhancement produces no measurable drop relative to the full method, the claim that these components are necessary would be falsified.

Figures

Figures reproduced from arXiv: 2604.05687 by Fei Wang, Jiaqi Zhao, Junjie Chen, Kun Li, Xinye Zheng, Yanyan Wei, Yiqi Nie, Zhiliang Wu.

Figure 1
Figure 1. Figure 1: Overview of our Smoke-GS method for hazy image restoration and 3D reconstruction. The pipeline begins with a hazy image [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
read the original abstract

Reconstructing 3D scenes from smoke-degraded multi-view images is particularly difficult because smoke introduces strong scattering effects, view-dependent appearance changes, and severe degradation of cross-view consistency. To address these issues, we propose a framework that integrates visual priors with efficient 3D scene modeling. We employ Nano-Banana-Pro to enhance smoke-degraded images and provide clearer visual observations for reconstruction and develop Smoke-GS, a medium-aware 3D Gaussian Splatting framework for smoke scene reconstruction and restoration-oriented novel view synthesis. Smoke-GS models the scene using explicit 3D Gaussians and introduces a lightweight view-dependent medium branch to capture direction-dependent appearance variations caused by smoke. Our method preserves the rendering efficiency of 3D Gaussian Splatting while improving robustness to smoke-induced degradation. Results demonstrate the effectiveness of our method for generating consistent and visually clear novel views in challenging smoke environments.

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.

Referee Report

2 major / 1 minor

Summary. The paper claims that enhancing smoke-degraded multi-view images with Nano-Banana-Pro (an MLLM) and modeling the scene via Smoke-GS—a 3D Gaussian Splatting extension that adds a lightweight view-dependent medium branch—enables robust 3D reconstruction and consistent, clear novel-view synthesis in the presence of smoke scattering and degradation, while preserving the efficiency of standard 3DGS.

Significance. If the central claims were supported by evidence, the work would offer a practical, efficient route to 3D scene recovery under strong scattering by combining MLLM priors with an explicit medium-aware representation; this could benefit downstream tasks such as robotics or surveillance in adverse environments. The emphasis on retaining real-time rendering speed is a clear strength.

major comments (2)
  1. Abstract: the statement that 'Results demonstrate the effectiveness of our method' is unsupported because the manuscript supplies no quantitative metrics, ablation studies, error bars, datasets, baselines, or experimental protocol, rendering it impossible to verify whether the MLLM enhancement or the medium branch actually improves reconstruction or novel-view consistency.
  2. Smoke-GS framework description: the lightweight view-dependent medium branch is introduced to 'capture direction-dependent appearance variations caused by smoke' yet no details are given on its parameterization, auxiliary losses, density regularization, multi-view medium consistency constraints, or any term enforcing energy conservation or reciprocity; without these, the branch can absorb residual appearance errors rather than model the underlying 3D scattering process, undermining the claim of true medium-aware reconstruction.
minor comments (1)
  1. The terms 'Nano-Banana-Pro' and 'Smoke-GS' should be defined on first use and their relationship to existing MLLMs or 3DGS variants clarified.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and constructive review. The comments highlight important areas where the manuscript can be strengthened, particularly regarding experimental validation and technical details of the proposed framework. We address each major comment below and will incorporate revisions to improve clarity and rigor.

read point-by-point responses

  1. Referee: Abstract: the statement that 'Results demonstrate the effectiveness of our method' is unsupported because the manuscript supplies no quantitative metrics, ablation studies, error bars, datasets, baselines, or experimental protocol, rendering it impossible to verify whether the MLLM enhancement or the medium branch actually improves reconstruction or novel-view consistency.

    Authors: We acknowledge the referee's concern that the abstract's claim requires stronger backing. The manuscript currently emphasizes the methodological novelty and provides qualitative visual results in the experiments section to illustrate improvements in novel-view synthesis under smoke. To address this directly, we will revise the abstract for precision and expand the experimental section with quantitative metrics (e.g., PSNR, SSIM), ablation studies isolating the MLLM enhancement and medium branch contributions, error bars, dataset descriptions, baselines, and full experimental protocols. These additions will enable verification of the claimed benefits. revision: yes

  2. Referee: Smoke-GS framework description: the lightweight view-dependent medium branch is introduced to 'capture direction-dependent appearance variations caused by smoke' yet no details are given on its parameterization, auxiliary losses, density regularization, multi-view medium consistency constraints, or any term enforcing energy conservation or reciprocity; without these, the branch can absorb residual appearance errors rather than model the underlying 3D scattering process, undermining the claim of true medium-aware reconstruction.

    Authors: We agree that the current description of the view-dependent medium branch lacks sufficient technical specificity. In the revised manuscript, we will expand Section 3.2 to detail the branch's parameterization (as a compact MLP taking view direction and Gaussian attributes), the auxiliary loss terms (including reconstruction, regularization on density, and multi-view consistency penalties), and how the design approximates scattering effects while promoting physical plausibility through implicit constraints on appearance variation. This will better distinguish modeling of the medium from error absorption and support the medium-aware reconstruction claim. revision: yes

Circularity Check

0 steps flagged

No circularity: additive empirical framework with no load-bearing derivations or self-referential reductions

full rationale

The paper describes an empirical pipeline: MLLM-based image enhancement followed by a modified 3D Gaussian Splatting model with an added lightweight view-dependent medium branch. No equations, uniqueness theorems, or first-principles derivations are presented that reduce claimed performance to fitted parameters by construction or to self-citations. The medium branch is introduced as a modeling choice whose parameters are optimized via standard image reconstruction losses; this is an additive architectural decision rather than a self-definitional or fitted-input prediction. The central claims rest on experimental results in smoke environments, not on any chain that collapses to the inputs. This is the most common honest non-finding for applied CV reconstruction papers.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 2 invented entities

Review performed on abstract only; no explicit free parameters, mathematical axioms, or independent evidence for new entities can be extracted or audited.

invented entities (2)
  • Smoke-GS no independent evidence
    purpose: medium-aware 3D Gaussian Splatting framework for smoke scenes
    New modeling component introduced to handle view-dependent smoke effects.
  • Nano-Banana-Pro no independent evidence
    purpose: enhance smoke-degraded images to provide clearer observations
    External MLLM employed as the source of vision priors.

pith-pipeline@v0.9.0 · 5472 in / 1246 out tokens · 48398 ms · 2026-05-10T19:40:15.615362+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 7 Pith papers

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

  1. Dehaze-then-Splat: Generative Dehazing with Physics-Informed 3D Gaussian Splatting for Smoke-Free Novel View Synthesis

    cs.CV 2026-04 unverdicted novelty 5.0

    Dehaze-then-Splat uses per-frame generative dehazing followed by physics-regularized 3D Gaussian Splatting to achieve 20.98 dB PSNR and 0.683 SSIM on the Akikaze scene, a 1.5 dB gain over baseline by mitigating cross-...

  2. CLIP-Guided Data Augmentation for Night-Time Image Dehazing

    cs.CV 2026-04 unverdicted novelty 5.0

    CLIP-guided selection of external data plus staged NAFNet training and inference fusion provides an effective pipeline for nighttime image dehazing in the NTIRE 2026 challenge.

  3. Training-Free Model Ensemble for Single-Image Super-Resolution via Strong-Branch Compensation

    cs.CV 2026-04 unverdicted novelty 4.0

    A dual-branch training-free ensemble fuses a hybrid attention network with a Mamba-based model via weighted combination to enhance super-resolution PSNR on DIV2K x4.

  4. Dual-Branch Remote Sensing Infrared Image Super-Resolution

    cs.CV 2026-04 unverdicted novelty 4.0

    Dual-branch fusion of HAT-L and MambaIRv2-L with eight-way ensemble and equal-weight averaging outperforms single branches on PSNR, SSIM, and challenge score for infrared super-resolution.

  5. SmokeGS-R: Physics-Guided Pseudo-Clean 3DGS for Real-World Multi-View Smoke Restoration

    cs.CV 2026-04 conditional novelty 4.0

    SmokeGS-R uses refined dark channel prior for pseudo-clean supervision to train 3DGS geometry, followed by ensemble-based appearance harmonization, achieving PSNR 15.21 and outperforming baselines on smoke restoration...

  6. Beyond Model Design: Data-Centric Training and Self-Ensemble for Gaussian Color Image Denoising

    cs.CV 2026-04 unverdicted novelty 3.0

    Expanding training data diversity, adopting two-stage optimization, and applying geometric self-ensemble raises Restormer performance on Gaussian color denoising at sigma=50 by 3.366 dB PSNR on the NTIRE 2026 validation set.

  7. NTIRE 2026 3D Restoration and Reconstruction in Real-world Adverse Conditions: RealX3D Challenge Results

    cs.CV 2026-04 unverdicted novelty 2.0

    The NTIRE 2026 challenge reports measurable progress in 3D reconstruction pipelines that handle real-world low-light and smoke degradation via the RealX3D benchmark.

Reference graph

Works this paper leans on

61 extracted references · 20 canonical work pages · cited by 7 Pith papers · 12 internal anchors

  1. [1]

    eDiff-I: Text-to-Image Diffusion Models with an Ensemble of Expert Denoisers

    Yogesh Balaji, Seungjun Nah, et al. ediff-i: Text-to-image diffusion models with an ensemble of expert denoisers.arXiv preprint arXiv:2211.01324, 2022. 2

    work page internal anchor Pith review arXiv 2022

  2. [2]

    In- structpix2pix: Learning to follow image editing instructions

    Tim Brooks, Aleksander Holynski, and Alexei A Efros. In- structpix2pix: Learning to follow image editing instructions. InProceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 18392–18402, 2023. 2

    2023

  3. [3]

    Dehazenet: An end-to-end system for single image haze removal.IEEE transactions on image process- ing, 25(11):5187–5198, 2016

    Bolun Cai, Xiangmin Xu, Kui Jia, Chunmei Qing, and Dacheng Tao. Dehazenet: An end-to-end system for single image haze removal.IEEE transactions on image process- ing, 25(11):5187–5198, 2016. 1

    2016

  4. [4]

    GenSmoke-GS: A Multi-Stage Method for Novel View Synthesis from Smoke-Degraded Images Using a Generative Model

    Qida Cao, Xinyuan Hu, Changyue Shi, Jiajun Ding, Zhou Yu, and Jun Yu. Gensmoke-gs: A multi-stage method for novel view synthesis from smoke-degraded images using a generative model.arXiv preprint arXiv:2604.03039, 2026. 1

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  5. [5]

    Beyond Model Design: Data-Centric Training and Self-Ensemble for Gaussian Color Image Denoising

    Gengjia Chang, Xining Ge, Weijun Yuan, Zhan Li, Qiurong Song, Luen Zhu, and Shuhong Liu. Beyond model design: Data-centric training and self-ensemble for gaussian color image denoising.arXiv preprint arXiv:2604.11468, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  6. [6]

    Training-Free Model Ensemble for Single-Image Super-Resolution via Strong-Branch Compensation

    Gengjia Chang, Xining Ge, Weijun Yuan, Zhan Li, Qiurong Song, Luen Zhu, and Shuhong Liu. Training-free model en- semble for single-image super-resolution via strong-branch compensation.arXiv preprint arXiv:2604.11564, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  7. [7]

    Towards seamless interaction: Causal turn-level modeling of inter- active 3d conversational head dynamics.arXiv preprint arXiv:2512.15340, 2025

    Junjie Chen, Fei Wang, Zhihao Hunag, Qing Zhou, Kun Li, Dan Guo, Linfeng Zhang, and Xun Yang. Towards seamless interaction: Causal turn-level modeling of inter- active 3d conversational head dynamics.arXiv preprint arXiv:2512.15340, 2025. 1

    work page arXiv 2025

  8. [8]

    Snowformer: Context interaction transformer with scale- awareness for single image desnowing.arXiv preprint arXiv:2208.09703, 2022

    Sixiang Chen, Tian Ye, Yun Liu, and Erkang Chen. Snowformer: Context interaction transformer with scale- awareness for single image desnowing.arXiv preprint arXiv:2208.09703, 2022. 1

    work page arXiv 2022

  9. [9]

    Sparse sampling transformer with uncertainty- driven ranking for unified removal of raindrops and rain streaks

    Sixiang Chen, Tian Ye, Jinbin Bai, Erkang Chen, Jun Shi, and Lei Zhu. Sparse sampling transformer with uncertainty- driven ranking for unified removal of raindrops and rain streaks. InProceedings of the IEEE/CVF international con- ference on computer vision, pages 13106–13117, 2023. 1

    2023

  10. [10]

    Teaching tailored to talent: Adverse weather restoration via prompt pool and depth-anything con- straint

    Sixiang Chen, Tian Ye, Kai Zhang, Zhaohu Xing, Yunlong Lin, and Lei Zhu. Teaching tailored to talent: Adverse weather restoration via prompt pool and depth-anything con- straint. InEuropean Conference on Computer Vision, pages 95–115. Springer, 2024. 2

    2024

  11. [11]

    Dehazenerf: Multi-image haze removal and 3d shape reconstruction using neural radiance fields

    Wei-Ting Chen, Wang Yifan, Sy-Yen Kuo, and Gordon Wet- zstein. Dehazenerf: Multi-image haze removal and 3d shape reconstruction using neural radiance fields. In2024 Inter- national Conference on 3D Vision (3DV), pages 247–256. IEEE, 2024. 2

    2024

  12. [12]

    Dehaze-then-Splat: Generative Dehazing with Physics-Informed 3D Gaussian Splatting for Smoke-Free Novel View Synthesis

    Yuchao Chen and Hanqing Wang. Dehaze-then-splat: Gen- erative dehazing with physics-informed 3d gaussian splat- ting for smoke-free novel view synthesis.arXiv preprint arXiv:2604.13589, 2026. 1

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  13. [13]

    Focal network for image restoration

    Yuning Cui, Wenqi Ren, Xiaochun Cao, and Alois Knoll. Focal network for image restoration. InProceedings of the IEEE/CVF international conference on computer vision, pages 13001–13011, 2023. 1

    2023

  14. [14]

    Multi-scale boosted de- hazing network with dense feature fusion

    Hang Dong, Jinshan Pan, Lei Xiang, Zhe Hu, Xinyi Zhang, Fei Wang, and Ming-Hsuan Yang. Multi-scale boosted de- hazing network with dense feature fusion. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2157–2167, 2020. 1

    2020

  15. [15]

    SmokeGS-R: Physics-Guided Pseudo-Clean 3DGS for Real-World Multi-View Smoke Restoration

    Xueming Fu and Lixia Han. Smokegs-r: Physics-guided pseudo-clean 3dgs for real-world multi-view smoke restora- tion.arXiv preprint arXiv:2604.05301, 2026. 1

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  16. [16]

    Dual-Branch Remote Sensing Infrared Image Super-Resolution

    Xining Ge, Gengjia Chang, Weijun Yuan, Zhan Li, Zhanglu Chen, Boyang Yao, Yihang Chen, Yifan Deng, and Shuhong Liu. Dual-branch remote sensing infrared image super- resolution.arXiv preprint arXiv:2604.10112, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  17. [17]

    CLIP-Guided Data Augmentation for Night-Time Image Dehazing

    Xining Ge, Weijun Yuan, Gengjia Chang, Xuyang Li, and Shuhong Liu. Clip-guided data augmentation for night-time image dehazing.arXiv preprint arXiv:2604.05500, 2026. 1

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  18. [18]

    Aquanerf: Neural radiance fields in under- water media with distractor removal

    Luca Gough, Adrian Azzarelli, Fan Zhang, and Nantheera Anantrasirichai. Aquanerf: Neural radiance fields in under- water media with distractor removal. In2025 IEEE Inter- national Symposium on Circuits and Systems (ISCAS), pages 1–5. IEEE, 2025. 2

    2025

  19. [19]

    Benchmarking micro-action recognition: Dataset, methods, and applications.IEEE Transactions on Circuits and Systems for Video Technology, 34(7):6238–6252, 2024

    Dan Guo, Kun Li, Bin Hu, Yan Zhang, and Meng Wang. Benchmarking micro-action recognition: Dataset, methods, and applications.IEEE Transactions on Circuits and Systems for Video Technology, 34(7):6238–6252, 2024. 1

    2024

  20. [20]

    Reliability-aware staged low-light gaussian splatting.ResearchGate preprint, 2026

    Haojie Guo and Ke Xian. Reliability-aware staged low-light gaussian splatting.ResearchGate preprint, 2026. 1

    2026

  21. [21]

    Neuropump: Simultaneous geometric and color rectification for underwater images

    Yue Guo, Haoxiang Liao, Haibin Ling, and Bingyao Huang. Neuropump: Simultaneous geometric and color rectification for underwater images. InProceedings of the 33rd ACM International Conference on Multimedia, pages 422–431,

  22. [22]

    Optimizing prompts for text-to-image generation.Advances in Neural Information Processing Systems, 36:66923–66939, 2023

    Yaru Hao, Zewen Chi, Li Dong, and Furu Wei. Optimizing prompts for text-to-image generation.Advances in Neural Information Processing Systems, 36:66923–66939, 2023. 2

    2023

  23. [23]

    Unsupervised night image enhancement: When layer decomposition meets light-effects suppression

    Yeying Jin, Wenhan Yang, and Robby T Tan. Unsupervised night image enhancement: When layer decomposition meets light-effects suppression. InEuropean conference on com- puter vision, pages 404–421. Springer, 2022. 1

    2022

  24. [24]

    Multi-concept customization of text-to-image diffusion

    Nupur Kumari, Bingliang Zhang, Richard Zhang, Eli Shechtman, and Jun-Yan Zhu. Multi-concept customization of text-to-image diffusion. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1931–1941, 2023. 2

    1931

  25. [25]

    Watersplatting: Fast underwater 3d scene reconstruction using gaussian splatting

    Huapeng Li, Wenxuan Song, Tianao Xu, Alexandre Elsig, and Jonas Kulhanek. Watersplatting: Fast underwater 3d scene reconstruction using gaussian splatting. InInterna- tional Conference on 3D Vision, pages 969–978. IEEE, 2025. 2

    2025

  26. [26]

    Dehazing- nerf: neural radiance fields from hazy images.arXiv preprint arXiv:2304.11448, 2023

    Tian Li, LU Li, Wei Wang, and Zhangchi Feng. Dehazing- nerf: neural radiance fields from hazy images.arXiv preprint arXiv:2304.11448, 2023. 2

    work page arXiv 2023

  27. [27]

    Realx3d: A physically-degraded 3d benchmark for multi-view visual restoration and recon- struction.arXiv preprint arXiv:2512.23437, 2025

    Shuhong Liu, Chenyu Bao, Ziteng Cui, Yun Liu, Xuangeng Chu, Lin Gu, Marcos V Conde, Ryo Umagami, Tomohiro Hashimoto, Zijian Hu, et al. Realx3d: A physically-degraded 3d benchmark for multi-view visual restoration and recon- struction.arXiv preprint arXiv:2512.23437, 2025. 4 5

    work page arXiv 2025

  28. [28]

    I2-nerf: Learning neural radiance fields un- der physically-grounded media interactions

    Shuhong Liu, Lin Gu, Ziteng Cui, Xuangeng Chu, and Tat- suya Harada. I2-nerf: Learning neural radiance fields un- der physically-grounded media interactions. InAdvances in Neural Information Processing Systems (NeurIPS), 2025. 2

    2025

  29. [30]

    NTIRE 2026 3D Restoration and Reconstruction in Real-world Adverse Conditions: RealX3D Challenge Results

    Shuhong Liu, Chenyu Bao, Ziteng Cui, et al. Ntire 2026 3d restoration and reconstruction in real-world ad- verse conditions: Realx3d challenge results.arXiv preprint arXiv:2604.04135, 2026. 1

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  30. [31]

    ELoG-GS: Dual-Branch Gaussian Splatting with Luminance-Guided Enhancement for Extreme Low-light 3D Reconstruction

    Yuhao Liu, Dingju Wang, and Ziyang Zheng. Elog-gs: Dual- branch gaussian splatting with luminance-guided enhance- ment for extreme low-light 3d reconstruction.arXiv preprint arXiv:2604.12592, 2026. 1

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  31. [32]

    Dehazegs: 3d gaus- sian splatting for multi-image haze removal.IEEE Signal Processing Letters, 32:736–740, 2025

    Chenjun Ma, Jieyu Zhao, and Jian Chen. Dehazegs: 3d gaus- sian splatting for multi-image haze removal.IEEE Signal Processing Letters, 32:736–740, 2025. 2

    2025

  32. [33]

    T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models

    Chong Mou, Xintao Wang, Liangbin Xie, Yanze Wu, Jian Zhang, Zhongang Qi, and Ying Shan. T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models. InProceedings of the AAAI conference on artificial intelligence, pages 4296–4304, 2024. 2

    2024

  33. [34]

    One-step image translation with text-to-image models,

    Gaurav Parmar, Taesung Park, Srinivasa Narasimhan, and Jun-Yan Zhu. One-step image translation with text-to-image models.arXiv preprint arXiv:2403.12036, 2024. 2

    work page arXiv 2024

  34. [35]

    Cluster-phys: Facial clues clustering towards efficient re- mote physiological measurement

    Wei Qian, Kun Li, Dan Guo, Bin Hu, and Meng Wang. Cluster-phys: Facial clues clustering towards efficient re- mote physiological measurement. InProceedings of the 32nd ACM International Conference on Multimedia, pages 330– 339, 2024. 2

    2024

  35. [36]

    Joint spatial-temporal modeling and contrastive learning for self-supervised heart rate measure- ment.arXiv preprint arXiv:2406.04942, 2024

    Wei Qian, Qi Li, Kun Li, Xinke Wang, Xiao Sun, Meng Wang, and Dan Guo. Joint spatial-temporal modeling and contrastive learning for self-supervised heart rate measure- ment.arXiv preprint arXiv:2406.04942, 2024. 2

    work page arXiv 2024

  36. [37]

    Wei Qian, Gaoji Su, Dan Guo, Jinxing Zhou, Xiaobai Li, Bin Hu, Shengeng Tang, and Meng Wang. Physdiff: Physiology- based dynamicity disentangled diffusion model for remote physiological measurement.Proceedings of the AAAI Con- ference on Artificial Intelligence (AAAI), 2025. 2

    2025

  37. [38]

    Mb-taylorformer: Multi-branch efficient transformer expanded by taylor formula for image dehazing

    Yuwei Qiu, Kaihao Zhang, Chenxi Wang, Wenhan Luo, Hongdong Li, and Zhi Jin. Mb-taylorformer: Multi-branch efficient transformer expanded by taylor formula for image dehazing. InProceedings of the IEEE/CVF international conference on computer vision, pages 12802–12813, 2023. 2

    2023

  38. [39]

    Gendeg: Diffusion-based degradation synthesis for generalizable all-in-one image restoration

    Sudarshan Rajagopalan, Nithin Gopalakrishnan Nair, Jay N Paranjape, and Vishal M Patel. Gendeg: Diffusion-based degradation synthesis for generalizable all-in-one image restoration. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 28144– 28154, 2025. 2

    2025

  39. [40]

    Scattern- erf: Seeing through fog with physically-based inverse neural rendering

    Andrea Ramazzina, Mario Bijelic, Stefanie Walz, Alessan- dro Sanvito, Dominik Scheuble, and Felix Heide. Scattern- erf: Seeing through fog with physically-based inverse neural rendering. InProceedings of the IEEE/CVF international conference on computer vision, pages 17957–17968, 2023. 2

    2023

  40. [41]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 2

    2022

  41. [42]

    Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Yael Pritch, Michael Rubinstein, and Kfir Aberman. Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 22500– 22510, 2023. 2

    2023

  42. [43]

    Vision transformers for single image dehazing.IEEE Transactions on Image Processing, 32:1927–1941, 2023

    Yuda Song, Zhuqing He, Hui Qian, and Xin Du. Vision transformers for single image dehazing.IEEE Transactions on Image Processing, 32:1927–1941, 2023. 2

    1927

  43. [44]

    Neural underwater scene representation

    Yunkai Tang, Chengxuan Zhu, Renjie Wan, Chao Xu, and Boxin Shi. Neural underwater scene representation. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 11780–11789, 2024. 2

    2024

  44. [45]

    Eulermormer: Robust eulerian motion magnification via dynamic filtering within transformer

    Fei Wang, Dan Guo, Kun Li, and Meng Wang. Eulermormer: Robust eulerian motion magnification via dynamic filtering within transformer. InProceedings of the AAAI Conference on Artificial Intelligence, pages 5345–5353, 2024. 1

    2024

  45. [46]

    Frequency decoupling for motion magnification via multi-level isomorphic architecture

    Fei Wang, Dan Guo, Kun Li, Zhun Zhong, and Meng Wang. Frequency decoupling for motion magnification via multi-level isomorphic architecture. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18984–18994, 2024. 1

    2024

  46. [47]

    Xin- sight: Integrative stage-consistent psychological counsel- ing support agents for digital well-being.arXiv preprint arXiv:2603.06583, 2026

    Fei Wang, Jiangnan Yang, Junjie Chen, Yuxin Liu, Kun Li, Yanyan Wei, Dan Guo, and Meng Wang. Xin- sight: Integrative stage-consistent psychological counsel- ing support agents for digital well-being.arXiv preprint arXiv:2603.06583, 2026. 2

    work page arXiv 2026

  47. [48]

    Multimodal Protein Language Models for Enzyme Kinetic Parameters: From Substrate Recognition to Conformational Adaptation

    Fei Wang, Xinye Zheng, Kun Li, Yanyan Wei, Yuxin Liu, Ganpeng Hu, Tong Bao, and Jingwen Yang. Multimodal pro- tein language models for enzyme kinetic parameters: From substrate recognition to conformational adaptation.arXiv preprint arXiv:2603.12845, 2026. 2

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  48. [49]

    Task-generalized adaptive cross- domain learning for multimodal image fusion.IEEE Trans- actions on Multimedia, 2026

    Mengyu Wang, Zhenyu Liu, Kun Li, Yu Wang, Yuwei Wang, Yanyan Wei, and Fei Wang. Task-generalized adaptive cross- domain learning for multimodal image fusion.IEEE Trans- actions on Multimedia, 2026. 1

    2026

  49. [50]

    Low-light wheat image enhancement using an explicit inter-channel sparse transformer.Computers and Electronics in Agricul- ture, 224:109169, 2024

    Yu Wang, Fei Wang, Kun Li, Xuping Feng, Wenhui Hou, Lu Liu, Liqing Chen, Yong He, and Yuwei Wang. Low-light wheat image enhancement using an explicit inter-channel sparse transformer.Computers and Electronics in Agricul- ture, 224:109169, 2024. 1

    2024

  50. [51]

    Deraincyclegan: Rain attentive cyclegan for single image deraining and rain- making.IEEE Transactions on Image Processing, 30:4788– 4801, 2021

    Yanyan Wei, Zhao Zhang, Yang Wang, Mingliang Xu, Yi Yang, Shuicheng Yan, and Meng Wang. Deraincyclegan: Rain attentive cyclegan for single image deraining and rain- making.IEEE Transactions on Image Processing, 30:4788– 4801, 2021. 1

    2021

  51. [52]

    Leveraging vision-language prompts for real-world image restoration and enhancement.Com- 6 puter Vision and Image Understanding, 250:104222, 2025

    Yanyan Wei, Yilin Zhang, Kun Li, Fei Wang, Shengeng Tang, and Zhao Zhang. Leveraging vision-language prompts for real-world image restoration and enhancement.Com- 6 puter Vision and Image Understanding, 250:104222, 2025. 1

    2025

  52. [53]

    Plenodium: Underwater 3d scene reconstruc- tion with plenoptic medium representation.arXiv preprint arXiv:2505.21258, 2025

    Changguanng Wu, Jiangxin Dong, Chengjian Li, and Jin- hui Tang. Plenodium: Underwater 3d scene reconstruc- tion with plenoptic medium representation.arXiv preprint arXiv:2505.21258, 2025. 2

    work page arXiv 2025

  53. [54]

    Bvinet: Unlocking blind video inpainting with zero annota- tions

    Zhiliang Wu, Kerui Chen, Kun Li, Hehe Fan, and Yi Yang. Bvinet: Unlocking blind video inpainting with zero annota- tions. InProceedings of the IEEE/CVF International Con- ference on Computer Vision, pages 14017–14027, 2025. 1

    2025

  54. [55]

    Drafting and revision: advancing high-fidelity video inpainting

    Zhiliang Wu, Kun Li, Hehe Fan, and Yi Yang. Drafting and revision: advancing high-fidelity video inpainting. InPro- ceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence, pages 2063–2071, 2025

    2063

  55. [56]

    Dlvinet: Advancing dual-lens video inpainting beyond par- allax constraints

    Zhiliang Wu, Kun Li, Yunqiu Xu, Hehe Fan, and Yi Yang. Dlvinet: Advancing dual-lens video inpainting beyond par- allax constraints. InProceedings of the AAAI Conference on Artificial Intelligence, pages 10888–10896, 2026. 1

    2026

  56. [57]

    Imagere- ward: Learning and evaluating human preferences for text- to-image generation.Advances in Neural Information Pro- cessing Systems, 36:15903–15935, 2023

    Jiazheng Xu, Xiao Liu, Yuchen Wu, Yuxuan Tong, Qinkai Li, Ming Ding, Jie Tang, and Yuxiao Dong. Imagere- ward: Learning and evaluating human preferences for text- to-image generation.Advances in Neural Information Pro- cessing Systems, 36:15903–15935, 2023. 2

    2023

  57. [58]

    Sea- splat: Representing underwater scenes with 3d gaussian splatting and a physically grounded image formation model

    Daniel Yang, John J Leonard, and Yogesh Girdhar. Sea- splat: Representing underwater scenes with 3d gaussian splatting and a physically grounded image formation model. InIEEE International Conference on Robotics and Automa- tion, pages 7632–7638. IEEE, 2025. 2

    2025

  58. [59]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF international conference on computer vision, pages 3836–3847, 2023. 2

    2023

  59. [60]

    Decoupling scattering: Pseudo-label guided nerf for scenes with scattering media

    Mingyang Zhang, Junkang Zhang, Faming Fang, and Guixu Zhang. Decoupling scattering: Pseudo-label guided nerf for scenes with scattering media. InProceedings of the AAAI Conference on Artificial Intelligence, pages 10031–10039,

  60. [61]

    Tg4mm: Time-varying gaussian splatting for 3d motion magnification.IEEE Transactions on Circuits and Systems for Video Technology, 2026

    Zheng Zhang, Jiabao Guo, Fei Wang, Jinyang Huang, Zhi Liu, and Dan Guo. Tg4mm: Time-varying gaussian splatting for 3d motion magnification.IEEE Transactions on Circuits and Systems for Video Technology, 2026. 2

    2026

  61. [62]

    Naka-GS: A Bionics-inspired Dual-Branch Naka Correction and Progressive Point Pruning for Low-Light 3DGS

    Runyu Zhu, SiXun Dong, Zhiqiang Zhang, Qingxia Ye, and Zhihua Xu. Naka-gs: A bionics-inspired dual-branch naka correction and progressive point pruning for low-light 3dgs. arXiv preprint arXiv:2604.11142, 2026. 1 7

    work page internal anchor Pith review Pith/arXiv arXiv 2026