Pith. sign in

REVIEW 18 cited by

3D Gaussian Splatting as Markov Chain Monte Carlo

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2404.09591 v3 pith:TOAZCKMD submitted 2024-04-15 cs.CV

classification cs.CV
keywords gaussiansgaussiansplattingcarlochaincloninginitializationmarkov
verification ladder T0 review T1 audit T2 compute T3 formal

Signed reviews

No signed human review yet.

0 comments
read the original abstract

While 3D Gaussian Splatting has recently become popular for neural rendering, current methods rely on carefully engineered cloning and splitting strategies for placing Gaussians, which can lead to poor-quality renderings, and reliance on a good initialization. In this work, we rethink the set of 3D Gaussians as a random sample drawn from an underlying probability distribution describing the physical representation of the scene-in other words, Markov Chain Monte Carlo (MCMC) samples. Under this view, we show that the 3D Gaussian updates can be converted as Stochastic Gradient Langevin Dynamics (SGLD) updates by simply introducing noise. We then rewrite the densification and pruning strategies in 3D Gaussian Splatting as simply a deterministic state transition of MCMC samples, removing these heuristics from the framework. To do so, we revise the 'cloning' of Gaussians into a relocalization scheme that approximately preserves sample probability. To encourage efficient use of Gaussians, we introduce a regularizer that promotes the removal of unused Gaussians. On various standard evaluation scenes, we show that our method provides improved rendering quality, easy control over the number of Gaussians, and robustness to initialization.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 18 Pith papers

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

  1. ContraGS: Codebook-Condensed and Trainable Gaussian Splatting for Fast, Memory-Efficient Reconstruction

    cs.GR 2025-09 reject novelty 7.0 of 10

    ContraGS trains 3D Gaussian Splatting directly on codebook-compressed representations, cutting peak model memory ~3.5x with small quality loss.

  2. LSD-3D: Large-Scale 3D Driving Scene Generation with Geometry Grounding

    cs.CV 2025-08 conditional novelty 7.0 of 10

    LSD-3D generates explicit, 3D-consistent driving scenes by combining a generated proxy mesh with geometry-grounded distillation from a 2D diffusion model.

  3. Gaussian Splatting for Efficient Satellite Image Photogrammetry

    cs.CV 2024-12 conditional novelty 7.0 of 10

    A Gaussian-splatting pipeline with affine satellite cameras, shadow mapping, and three regularizers reconstructs terrain with accuracy close to EO-NeRF in about 300x less training time.

  4. SplatCtrl: Perception-Action Coupling via Gaussian Scene Representations and Reactive Robot Control

    cs.RO 2026-07 conditional novelty 6.0 of 10

    SplatCtrl couples real-time isotropic Gaussian scene reconstruction from RGB-D with continuous GPDF-derived SDFs inside control-barrier QP-IK for collision-free 6-DoF robot motion in dynamic environments.

  5. PixGS: Pixel-Space Diffusion for Direct 3D Gaussian Splat Generation

    cs.CV 2026-07 unverdicted novelty 6.0 of 10

    PixGS is a single-stage pixel-space diffusion model that directly produces high-quality 3D Gaussian Splats from text or images in ~1s, outperforming multi-stage latent methods on standard benchmarks.

  6. CF3: Compact and Fast 3D Feature Fields

    cs.CV 2025-08 conditional novelty 6.0 of 10

    CF3 builds a compact 3D feature field from a pre-trained 3DGS by feature lifting, per-Gaussian autoencoding, and adaptive sparsification, matching baseline segmentation quality with roughly 5% of the Gaussians.

  7. 3D Scene-Camera Representation with Joint Camera Photometric Optimization

    cs.CV 2025-06 conditional novelty 6.0 of 10

    A 3D Gaussian splatting pipeline jointly optimizes a per-pixel photometric camera MLP and the scene radiance, removing vignetting, dirt, and defocus during novel-view synthesis.

  8. Joint Optimization of Neural Radiance Fields and Continuous Camera Motion from a Monocular Video

    cs.CV 2025-04 conditional novelty 6.0 of 10

    A NeRF-based system estimates camera motion as continuous velocities and recovers accurate camera poses and scene geometry from monocular video without depth priors.

  9. Casual3DHDR: Deblurring High Dynamic Range 3D Gaussian Splatting from Casually Captured Videos

    cs.CV 2025-04 conditional novelty 6.0 of 10

    A one-stage pipeline jointly estimates camera motion, exposure times, and the camera response while reconstructing an HDR 3D scene from blurry auto-exposure videos.

  10. Learning Radiance Fields from a Single Snapshot Compressive Image

    cs.CV 2024-12 conditional novelty 6.0 of 10

    SCINeRF and SCISplat recover a view-consistent 3D scene representation from a single snapshot compressive image, with SCISplat reaching 35.94 dB PSNR and 205 FPS rendering.

  11. 3DGUT: Enabling Distorted Cameras and Secondary Rays in Gaussian Splatting

    cs.GR 2024-12 conditional novelty 6.0 of 10

    Replacing EWA splatting's linearized projection with an Unscented Transform lets 3DGS handle fisheye and rolling-shutter cameras and enables hybrid rasterization plus traced secondary rays.

  12. Generative Densification: Learning to Densify Gaussians for High-Fidelity Generalizable 3D Reconstruction

    cs.CV 2024-12 conditional novelty 6.0 of 10

    Generative Densification improves feed-forward Gaussian 3D reconstruction by learning to generate fine Gaussians for detailed regions in one forward pass, and it beats baselines on object and scene datasets.

  13. Volumetrically Consistent 3D Gaussian Rasterization

    cs.CV 2024-12 conditional novelty 6.0 of 10

    A drop-in alpha computation for 3DGS rasterizers, derived from analytic volumetric transmittance, improves edge and perceptual metrics and transfers to tomography.

  14. Dynamics-Aware Gaussian Splatting Streaming Towards Fast On-the-Fly 4D Reconstruction

    cs.CV 2024-11 conditional novelty 6.0 of 10

    A dynamics-aware three-stage Gaussian splatting pipeline achieves the fastest reported on-the-fly 4D reconstruction training with competitive quality on N3DV and MeetRoom benchmarks.

  15. Adam SLAM - the last mile of camera calibration with 3DGS

    cs.CV 2025-08 conditional novelty 5.0 of 10

    Backpropagating the 3D Gaussian Splatting color loss into camera parameters, with an L2 loss, focal-length training, and a Hessian-based reparameterization, raises average PSNR by 0.4 dB over COLMAP calibration on ref...

  16. GSCodec Studio: A Modular Framework for Gaussian Splat Compression

    cs.CV 2025-06 conditional novelty 5.0 of 10

    GSCodec Studio is a modular open-source framework for Gaussian Splat compression, and its composed Static and Dynamic GSCodec pipelines report competitive rate-distortion results against several baselines.

  17. Steepest Descent Density Control for Compact 3D Gaussian Splatting

    cs.CV 2025-05 conditional novelty 5.0 of 10

    SteepGS splits a 3D Gaussian only when a computed splitting matrix has a negative eigenvalue, placing two half-opacity offspring along the steepest descent direction, achieving about 50% point reduction with comparabl...

  18. Gaussian Object Carver: Object-Compositional Gaussian Splatting with surfaces completion

    cs.CV 2024-12 conditional novelty 5.0 of 10

    A Gaussian splatting pipeline reconstructs indoor scenes as separable objects and uses a trained completion model to fill in occluded surfaces zero-shot.

Pith tools