mlsanityData to 14 Oct 2026

RealX3D: A Physically-Degraded 3D Benchmark for Multi-view Visual Restoration and Reconstruction

arXiv 2512.23437 ↗First posted 29 Dec 2025cs.CV cs.MMcited by 19

Not on the map. Why?

Abstract (arXiv)

We introduce RealX3D, a real-capture benchmark for multi-view visual restoration and 3D reconstruction under diverse physical degradations. RealX3D groups corruptions into four families, including illumination, scattering, occlusion, and blurring, and captures each at multiple severity levels using a unified acquisition protocol that yields pixel-aligned LQ/GT views. Each scene includes high-resolution capture, RAW images, and dense laser scans, from which we derive world-scale meshes and metric depth. Benchmarking a broad range of optimization-based and feed-forward methods shows substantial degradation in reconstruction quality under physical corruptions, underscoring the fragility of current multi-view pipelines in real-world challenging environments.

Key facts. Cited by 19 papers in the map (data through 14 Oct 2026). Builds on 99 of its 132 references that are in the map. First posted on arXiv 29 Dec 2025.

Builds on

99 of 132 references.

References · 99

  1. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  2. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
  3. 20233D Gaussian Splatting for Real-Time Radiance Field Renderingcited by 6,226
  4. 2021Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fieldscited by 2,032
  5. 2023DUSt3R: Geometric 3D Vision Made Easycited by 1,567
  6. 2025VGGT: Visual Geometry Grounded Transformercited by 1,539
  7. 2020NeRF in the Wild: Neural Radiance Fields for Unconstrained Photo Collectionscited by 1,232
  8. 2025Depth Anything 3: Recovering the Visual Space from Any Viewscited by 692
  9. 2025Continuous 3D Perception Model with Persistent Statecited by 548
  10. 2021Neural Fields in Visual Computing and Beyondcited by 540
  11. 2024MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motioncited by 442
  12. 2025Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Passcited by 367
38 more
  1. 2021NeRF in the Dark: High Dynamic Range View Synthesis from Noisy Raw Imagescited by 336
  2. 2025π3: Permutation-Equivariant Visual Geometry Learningcited by 319
  3. 2025MapAnything: Universal Feed-Forward Metric 3D Reconstructioncited by 314
  4. 2021Deblur-NeRF: Neural Radiance Fields from Blurry Imagescited by 179
  5. 2025Difix3D+: Improving 3D Reconstructions with Single-Step Diffusion Modelscited by 161
  6. 2025Streaming 4D Visual Geometry Transformercited by 161
  7. 2024WildGaussians: 3D Gaussian Splatting in the Wildcited by 152
  8. 2022D2NeRF: Self-Supervised Decoupling of Dynamic and Static Objects from a Monocular Videocited by 137
  9. 2021Hallucinated Neural Radiance Fields in the Wildcited by 130
  10. 2025TTT3R: 3D Reconstruction as Test-Time Trainingcited by 129
  11. 2024Gaussian in the Wild: 3D Gaussian Splatting for Unconstrained Image Collectionscited by 117
  12. 2021HDR-NeRF: High Dynamic Range Neural Radiance Fieldscited by 114
  13. 2023RobustNeRF: Ignoring Distractors with Robust Lossescited by 101
  14. 2025Easi3R: Estimating Disentangled Motion from DUSt3R Without Trainingcited by 91
  15. 2025St4RTrack: Simultaneous 4D Reconstruction and Tracking in the Worldcited by 89
  16. 2024SGS-SLAM: Semantic Gaussian Splatting For Neural Dense SLAMcited by 87
  17. 2024NeRF On-the-go: Exploiting Uncertainty for Distractor-free NeRFs in the Wildcited by 85
  18. 2022BAD-NeRF: Bundle Adjusted Deblur Neural Radiance Fieldscited by 80
  19. 2023SeaThru-NeRF: Neural Radiance Fields in Scattering Mediacited by 75
  20. 2022DP-NeRF: Deblurred Neural Radiance Field with Physical Scene Priorscited by 70
  21. 2024Deblurring 3D Gaussian Splattingcited by 66
  22. 2024BAD-Gaussians: Bundle Adjusted Deblur Gaussian Splattingcited by 61
  23. 2024SpotlessSplats: Ignoring Distractors in 3D Gaussian Splattingcited by 61
  24. 2023Cross-Ray Neural Radiance Fields for Novel-view Synthesis from Unconstrained Image Collectionscited by 54
  25. 2024BAGS: Blur Agnostic Gaussian Splatting through Multi-Scale Kernel Modelingcited by 42
  26. 2024WaterSplatting: Fast Underwater 3D Scene Reconstruction Using Gaussian Splattingcited by 41
  27. 2023Lighting up NeRF via Unsupervised Decomposition and Enhancementcited by 40
  28. 2023ExBluRF: Efficient Radiance Fields for Extreme Motion Blurred Imagescited by 38
  29. 2024SeaSplat: Representing Underwater Scenes with 3D Gaussian Splatting and a Physically Grounded Image Formation Modelcited by 36
  30. 2025On-the-fly Reconstruction for Large-Scale Novel View Synthesis from Unposed Imagescited by 36
  31. 2023Aleth-NeRF: Illumination Adaptive NeRF with Concealing Field Assumptioncited by 35
  32. 2024HDR-GS: Efficient High Dynamic Range Novel View Synthesis at 1000x Speed via Gaussian Splattingcited by 35
  33. 2023NeRFLiX: High-Quality Neural View Synthesis by Learning a Degradation-Driven Inter-viewpoint MiXercited by 29
  34. 2023ScatterNeRF: Seeing Through Fog with Physically-Based Inverse Neural Renderingcited by 28
  35. 2024Bilateral Guided Radiance Field Processingcited by 28
  36. 2024Lighting Every Darkness with 3DGS: Fast Training and Real-Time Rendering for HDR View Synthesiscited by 27
  37. 2025Luminance-GS: Adapting 3D Gaussian Splatting to Challenging Lighting Conditions with View-Adaptive Curve Adjustmentcited by 27
  38. 2024MG-SLAM: Structure Gaussian Splatting SLAM with Manhattan World Hypothesiscited by 22

50 most cited of 99 All 99 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 19.

19 papers.

Sort:

By area:

  1. 2026GenSmoke-GS: A Multi-Stage Method for Novel View Synthesis from Smoke-Degraded Images Using a Generative Modelcited by 11
  2. 2026NTIRE 2026 3D Restoration and Reconstruction in Real-world Adverse Conditions: RealX3D Challenge Resultscited by 11
  3. 2026Unifying Color and Lightness Correction with View-Adaptive Curve Adjustment for Robust 3D Novel View Synthesiscited by 8
  4. 2026CLIP-Guided Data Augmentation for Night-Time Image Dehazingcited by 8
  5. 20263D Smoke Scene Reconstruction Guided by Vision Priors from Multimodal Large Language Modelscited by 7
  6. 2026Dual-Branch Remote Sensing Infrared Image Super-Resolutioncited by 7
  7. 2026SmokeGS-R: Physics-Guided Pseudo-Clean 3DGS for Real-World Multi-View Smoke Restorationcited by 6
  8. 2026Training-Free Model Ensemble for Single-Image Super-Resolution via Strong-Branch Compensationcited by 6
  9. 2026Beyond Model Design: Data-Centric Training and Self-Ensemble for Gaussian Color Image Denoisingcited by 5
  10. 2026RAWild: Sensor-Agnostic RAW Object Detection via Physics-Guided Curve and Grid Modelingcited by 3
  11. 2026Naka-GS: Naka-Guided Chroma Correction and Point-Prior Compaction for Low-Light 3D Gaussian Splattingcited by 2
  12. 2026FluxFlow: Conservative Flow-Matching for Astronomical Image Super-Resolutioncited by 2
7 more
  1. 2026ELoG-GS: Dual-Branch Gaussian Splatting with Luminance-Guided Enhancement for Extreme Low-light 3D Reconstructioncited by 1
  2. 2026Dehaze-then-Splat: Generative Dehazing with Physics-Informed 3D Gaussian Splatting for Smoke-Free Novel View Synthesiscited by 1
  3. 2026DF3DV-1K: A Large-Scale Dataset and Benchmark for Distractor-Free Novel View Synthesiscited by 0
  4. 2026Scene-Adaptive Nonlinear Tone Curves for Pseudo Ground-Truth Generation in Low-Light 3D Gaussian Splattingcited by 0
  5. 2026BlindPSNR: A No-Reference Fidelity Predictor for Low-Light Image Enhancementcited by 0
  6. 2026FujinSplat: Seeing Through Smoke with RAW-Domain Gaussian Splattingcited by 0
  7. 2026ASTRA-SR: Atmospheric Seeing and Turbulence Restoration for Astronomical Image Super-Resolutioncited by 0