mlsanityData to 7 Oct 2026

AirSplat: Alignment and Rating for Robust Feed-Forward 3D Gaussian Splatting

arXiv 2603.25129 ↗First posted 26 Mar 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

While 3D Vision Foundation Models (3DVFMs) have demonstrated remarkable zero-shot capabilities in visual geometry estimation, their direct application to generalizable novel view synthesis (NVS) remains challenging. In this paper, we propose AirSplat, a novel training framework that effectively adapts the robust geometric priors of 3DVFMs into high-fidelity, pose-free NVS. Our approach introduces two key technical contributions: (1) Self-Consistent Pose Alignment (SCPA), a training-time feedback loop that ensures pixel-aligned supervision to resolve pose-geometry discrepancy; and (2) Rating-based Opacity Matching (ROM), which leverages the local 3D geometry consistency knowledge from a sparse-view NVS teacher model to filter out degraded primitives. Experimental results on large-scale benchmarks demonstrate that our method significantly outperforms state-of-the-art pose-free NVS approaches in reconstruction quality. Our AirSplat highlights the potential of adapting 3DVFMs to enable simultaneous visual geometry estimation and high-quality view synthesis.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 55 of its 69 references that are in the map. First posted on arXiv 26 Mar 2026.

Builds on

55 of 69 references.

References · 55

  1. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  2. 20233D Gaussian Splatting for Real-Time Radiance Field Renderingcited by 6,141
  3. 2022Instant Neural Graphics Primitives with a Multiresolution Hash Encodingcited by 3,698
  4. 2021Plenoxels: Radiance Fields without Neural Networkscited by 1,607
  5. 2020pixelNeRF: Neural Radiance Fields from One or Few Imagescited by 1,574
  6. 2023DUSt3R: Geometric 3D Vision Made Easycited by 1,526
  7. 2025VGGT: Visual Geometry Grounded Transformercited by 1,475
  8. 2022TensoRF: Tensorial Radiance Fieldscited by 1,275
  9. 20234D Gaussian Splatting for Real-Time Dynamic Scene Renderingcited by 1,144
  10. 20242D Gaussian Splatting for Geometrically Accurate Radiance Fieldscited by 1,061
  11. 2024Grounding Image Matching in 3D with MASt3Rcited by 960
  12. 2021PlenOctrees for Real-time Rendering of Neural Radiance Fieldscited by 904
38 more
  1. 2021MVSNeRF: Fast Generalizable Radiance Field Reconstruction from Multi-View Stereocited by 735
  2. 2021IBRNet: Learning Multi-View Image-Based Renderingcited by 698
  3. 2023pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstructioncited by 668
  4. 2023Mip-Splatting: Alias-free 3D Gaussian Splattingcited by 661
  5. 2025Depth Anything 3: Recovering the Visual Space from Any Viewscited by 626
  6. 2024MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-View Imagescited by 553
  7. 2023DL3DV-10K: A Large-Scale Scene Dataset for Deep Learning-based 3D Visioncited by 514
  8. 2023Splatter Image: Ultra-Fast Single-View 3D Reconstructioncited by 359
  9. 2025π3: Permutation-Equivariant Visual Geometry Learningcited by 292
  10. 2024DepthSplat: Connecting Gaussian Splatting and Depthcited by 275
  11. 2024No Pose, No Problem: Surprisingly Simple 3D Gaussian Splats from Sparse Unposed Imagescited by 241
  12. 2020Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Imagecited by 239
  13. 2025AnySplat: Feed-forward 3D Gaussian Splatting from Unconstrained Viewscited by 234
  14. 2023COLMAP-Free 3D Gaussian Splattingcited by 193
  15. 2025FLARE: Feed-forward Geometry, Appearance and Camera Estimation from Uncalibrated Sparse Viewscited by 176
  16. 2024Splatt3R: Zero-shot Gaussian Splatting from Uncalibrated Image Pairscited by 171
  17. 2017Interactive Learning from Policy-Dependent Human Feedbackcited by 155
  18. 2022Generalizable Patch-Based Neural Renderingcited by 121
  19. 2024Long-LRM: Long-sequence Large Reconstruction Model for Wide-coverage Gaussian Splatscited by 111
  20. 2025RayZer: A Self-supervised Large View Synthesis Modelcited by 74
  21. 2024InstantSplat: Sparse-view Gaussian Splatting in Secondscited by 67
  22. 2025WorldMirror: Universal 3D World Reconstruction with Any-Prior Promptingcited by 60
  23. 2019Deep Reinforcement Learning from Policy-Dependent Human Feedbackcited by 56
  24. 2024SplineGS: Robust Motion-Adaptive Spline for Real-Time Dynamic 3D Gaussians from Monocular Videocited by 44
  25. 2024SelfSplat: Pose-Free and 3D Prior-Free Generalizable 3D Gaussian Splattingcited by 40
  26. 2026NeoVerse: Enhancing 4D World Model with in-the-wild Monocular Videoscited by 39
  27. 2025No Pose at All: Self-Supervised Pose-Free 3D Gaussian Splatting from Sparse Viewscited by 38
  28. 2025YoNoSplat: You Only Need One Model for Feedforward 3D Gaussian Splattingcited by 34
  29. 2025ZPressor: Bottleneck-Aware Compression for Scalable Feed-Forward 3DGScited by 29
  30. 2025LongSplat: Robust Unposed 3D Gaussian Splatting for Casual Long Videoscited by 23
  31. 2025MonoSplat: Generalizable 3D Gaussian Splatting from Monocular Depth Foundation Modelscited by 15
  32. 2025Enhancing Rating-Based Reinforcement Learning to Effectively Leverage Feedback from Large Vision-Language Modelscited by 15
  33. 2023Rating-based Reinforcement Learningcited by 14
  34. 2025MoBGS: Motion Deblurring Dynamic 3D Gaussian Splatting for Blurry Monocular Videocited by 11
  35. 2025QuickSplat: Fast 3D Surface Reconstruction via Learned Gaussian Initializationcited by 11
  36. 2025TARO: Timestep-Adaptive Representation Alignment with Onset-Aware Conditioning for Synchronized Video-to-Audio Synthesiscited by 10
  37. 2024MDSGen: Fast and Efficient Masked Diffusion Temporal-Aware Transformers for Open-Domain Sound Generationcited by 9
  38. 2025One-Step Flow Q-Learning: Addressing the Diffusion Policy Bottleneck in Offline Reinforcement Learningcited by 9

50 most cited of 55 All 55 (JSON)

Built on by

Not cited yet.