Generative 3D Gaussians with Learned Density Control

• URL: http://arxiv.org/abs/2605.16355v1
• Date: Fri, 08 May 2026 17:54:25 GMT
• Title: Generative 3D Gaussians with Learned Density Control
• Abstract summary: We present Density-Sampled Gaussians (DeG), a novel 3D representation designed to bridge the gap between adaptive rendering primitives and scalable generative modeling.
• Score: 41.893507326094415
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We present Density-Sampled Gaussians (DeG), a novel 3D representation designed to bridge the gap between adaptive rendering primitives and scalable generative modeling. Unlike existing approaches that constrain 3D Gaussians to fixed voxel grids or arrays, DeG models Gaussian centers as samples from a learnable probability density function defined over an octree. This formulation provides a rigorous mathematical framework for adaptive density control: by jointly optimizing the spatial density and Gaussian attributes under rendering supervision, our model naturally concentrates primitives in regions of high geometric complexity. We achieve this via a new render loss contribution gradient that serves as a fully differentiable analogue to the discrete densification and pruning heuristics used in standard Gaussian Splatting. The resulting representation is highly flexible, supporting variable-resolution decoding from a single latent code by simply adjusting the sampling budget. To enable generative synthesis, we train a latent diffusion model on DeG. We identify a critical challenge in applying diffusion to unordered set-structured latents, which can significantly slow convergence, and propose VecSeq, a canonical re-indexing mechanism that anchors latent tokens to a deterministic 3D Sobol sequence. This transforms the ambiguous set-generation problem into a robust sequence modeling task. Extensive experiments demonstrate that our pipeline achieves state-of-the-art quality in single-image-to-3D generation, combining the structural adaptivity of unstructured primitives with the training stability of grid-based methods.

Related papers

• 3D Skew-Normal Splatting [36.91264690191168]
  3D Gaussian Splatting (3DGS) has emerged as a leading representation for real-time novel view synthesis.<n>We propose Skew-Normal Splatting (SNS), which adopts the Azzalini Skew-Normal distribution as the fundamental primitive.<n>SNS consistently improves reconstruction quality over Gaussian and recent non-Gaussian kernels.
  arXiv  Detail & Related papers  (2026-05-14T16:10:41Z)
• NG-GS: NeRF-Guided 3D Gaussian Splatting Segmentation [65.34304674634713]
  We introduce NG-GS, a framework for high-quality object segmentation in 3DGS that explicitly addresses boundary discretization.<n>We show that our method achieves photorealistic state-of-the-art performance, with significant gains in boundary mIoU.
  arXiv  Detail & Related papers  (2026-04-16T07:14:07Z)
• GaussianGPT: Towards Autoregressive 3D Gaussian Scene Generation [42.49842620609683]
  We introduce a transformer-based model that directly generates 3D Gaussians via next-token prediction.<n>The resulting tokens are serialized and modeled using a causal transformer with 3D rotary positional embedding.<n>Unlike diffusion-based methods that refine scenes holistically, our formulation constructs scenes step-by-step, naturally supporting completion, outpainting, controllable sampling via temperature, and flexible generation horizons.
  arXiv  Detail & Related papers  (2026-03-27T17:58:05Z)
• Joint Semantic and Rendering Enhancements in 3D Gaussian Modeling with Anisotropic Local Encoding [86.55824709875598]
  We propose a joint enhancement framework for 3D semantic Gaussian modeling that synergizes both semantic and rendering branches.<n>Unlike conventional point cloud shape encoding, we introduce an anisotropic 3D Gaussian Chebyshev descriptor to capture fine-grained 3D shape details.<n>We employ a cross-scene knowledge transfer module to continuously update learned shape patterns, enabling faster convergence and robust representations.
  arXiv  Detail & Related papers  (2026-01-05T18:33:50Z)
• Metropolis-Hastings Sampling for 3D Gaussian Reconstruction [31.840492077537018]
  We propose an adaptive sampling framework for 3D Gaussian Splatting (3DGS)<n>Our framework overcomes limitations by reformulating densification and pruning as a probabilistic sampling process.<n>Our approach achieves faster convergence while matching or modestly surpassing the view-synthesis quality of state-of-the-art models.
  arXiv  Detail & Related papers  (2025-06-15T19:12:37Z)
• GP-GS: Gaussian Processes for Enhanced Gaussian Splatting [15.263608848427136]
  This paper proposes a novel 3D reconstruction framework, Gaussian Processes enhanced Gaussian Splatting (GP-GS)<n>GP-GS enables adaptive and uncertainty-guided densification of sparse Structure-from-Motion point clouds.<n>Experiments conducted on synthetic and real-world datasets validate the effectiveness and practicality of the proposed framework.
  arXiv  Detail & Related papers  (2025-02-04T12:50:16Z)
• DiffGS: Functional Gaussian Splatting Diffusion [33.07847512591061]
  3D Gaussian Splatting (3DGS) has shown convincing performance in rendering speed and fidelity. However, the generation of Gaussian Splatting remains a challenge due to its discreteness and unstructured nature. We propose DiffGS, a general Gaussian generator based on latent diffusion models.
  arXiv  Detail & Related papers  (2024-10-25T16:08:08Z)
• L3DG: Latent 3D Gaussian Diffusion [74.36431175937285]
  L3DG is the first approach for generative 3D modeling of 3D Gaussians through a latent 3D Gaussian diffusion formulation. We employ a sparse convolutional architecture to efficiently operate on room-scale scenes. By leveraging the 3D Gaussian representation, the generated scenes can be rendered from arbitrary viewpoints in real-time.
  arXiv  Detail & Related papers  (2024-10-17T13:19:32Z)
• GaussianCube: A Structured and Explicit Radiance Representation for 3D Generative Modeling [55.05713977022407]
  We introduce a radiance representation that is both structured and fully explicit and thus greatly facilitates 3D generative modeling. We derive GaussianCube by first using a novel densification-constrained Gaussian fitting algorithm, which yields high-accuracy fitting. Experiments conducted on unconditional and class-conditioned object generation, digital avatar creation, and text-to-3D all show that our model synthesis achieves state-of-the-art generation results.
  arXiv  Detail & Related papers  (2024-03-28T17:59:50Z)
• Mesh-based Gaussian Splatting for Real-time Large-scale Deformation [58.18290393082119]
  It is challenging for users to directly deform or manipulate implicit representations with large deformations in the real-time fashion. We develop a novel GS-based method that enables interactive deformation. Our approach achieves high-quality reconstruction and effective deformation, while maintaining the promising rendering results at a high frame rate.
  arXiv  Detail & Related papers  (2024-02-07T12:36:54Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.