L2D2-GS: Learning to Densify for Feedforward Dynamic Gaussian Scene Reconstruction

• URL: http://arxiv.org/abs/2606.29374v1
• Date: Sun, 28 Jun 2026 12:51:17 GMT
• Title: L2D2-GS: Learning to Densify for Feedforward Dynamic Gaussian Scene Reconstruction
• Abstract summary: This paper presents L2D2-GS, a unified framework that reformulates generalizable reconstruction not as a one-shot regression, but as a robust iterative process of optimization and densification.<n>Our method achieves state-of-the-art reconstruction fidelity and strong zero-shot generalization, while using fewer primitives than competing baselines.
• Score: 45.197986759830314
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: High-fidelity reconstruction of dynamic urban environments is a cornerstone of autonomous driving simulation and large-scale world modeling. While 3D Gaussian Splatting (3DGS) has established a new standard for real-time rendering, its reliance on expensive per-scene optimization limits scalability. Conversely, recent feedforward methods that infer Gaussian parameters offer faster speed but face fundamental bottlenecks: they are memory-prohibitive at high resolutions and struggle to fuse dense multi-view observations consistently. This paper presents L2D2-GS, a unified framework that reformulates generalizable reconstruction not as a one-shot regression, but as a robust iterative process of optimization and densification. To resolve the ambiguity of supervision in primitive generation, we propose a self-supervised densification policy that derives explicit reward signals from global reconstruction gains to guide local densification. Furthermore, we mitigate irreversible early-stage artifacts through a geometric regularization mechanism, utilizing reparameterization to constrain the optimization manifold and prevent convergence to poor local optima. Extensive experiments on the PandaSet and Waymo datasets demonstrate that our method achieves state-of-the-art reconstruction fidelity and strong zero-shot generalization, while using fewer primitives than competing baselines.

Related papers

• CG-GLORE: A Conjugate Gradient-Based Global-Local Regularization Network for Sparse-View CT Reconstruction [3.7567584936415166]
  Sparse-view computed tomography (CT) reduces radiation dose by acquiring fewer projection views, but the resulting inverse problem is highly ill-posed.<n>Existing deep reconstruction methods have achieved promising performance, yet many rely on first-order updates or large regularization networks.<n>We propose textbfCG-GLORE, a compact deep unrolling framework inspired by second-order optimization for sparse-view CT reconstruction.
  arXiv  Detail & Related papers  (2026-08-15T14:08:44Z)
• EvTrajGS: Accurate and Efficient 3D Gaussian Splatting from Unposed Event Streams [108.45551580182384]
  EvTrajGS is an accurate and efficient 3D Gaussian Splatting framework for unposed event streams.<n>Our method enables reliable joint pose-scene optimization from coarse pose priors, eliminating the need for computationally expensive SLAM-style pipelines.
  arXiv  Detail & Related papers  (2026-08-09T09:00:30Z)
• Incremental Online Scene Reconstruction by 3D Gaussian Triangulation [22.398131643187384]
  We propose a novel online framework that incrementally reconstructs and updates high-fidelity explicit meshes.<n>We also present a direct meshing algorithm that efficiently extracts and updates the mesh from the Gaussian set.<n>Our framework significantly reduces memory and computational overhead during long-sequence processing.
  arXiv  Detail & Related papers  (2026-07-12T10:22:18Z)
• FlowObject: Flow Steering for Bridging Generative Priors and Reconstruction Fidelity [34.142288038459704]
  FlowObject is a framework that reformulates sparse-view 3D reconstruction as a training-free, guided inverse problem.<n>Our method significantly outperforms state-of-the-art generative models and optimization-based frameworks in both geometric completeness and view-dependent appearance fidelity.
  arXiv  Detail & Related papers  (2026-06-17T12:42:09Z)
• ERGO: Excess-Risk-Guided Optimization for High-Fidelity Monocular 3D Gaussian Splatting [63.138778159026934]
  We propose an adaptive optimization framework guided by excess risk decomposition, termed ERGO.<n> ERGO dynamically estimates the view-specific excess risk and adaptively adjust loss weights during optimization.<n>Experiments on the Google Scanned Objects dataset and the OmniObject3D dataset demonstrate the superiority of ERGO over existing state-of-the-art methods.
  arXiv  Detail & Related papers  (2026-02-10T20:44:43Z)
• Tessellation GS: Neural Mesh Gaussians for Robust Monocular Reconstruction of Dynamic Objects [41.62908563782648]
  3D Gaussian Splatting (GS) enables highly photorealistic scene reconstruction from posed image sequences but struggles with viewpoint extrapolation.<n>We propose Tessellation GS, a structured 2D GS approach anchored on mesh faces, to reconstruct dynamic scenes from a single continuously moving or static camera.<n>Our method constrains 2D Gaussians to localized regions and infers their attributes via hierarchical neural features on mesh faces.
  arXiv  Detail & Related papers  (2025-12-08T10:16:14Z)
• Re-Activating Frozen Primitives for 3D Gaussian Splatting [22.003309830812825]
  3D Gaussian Splatting (3D-GS) achieves real-time photorealistic novel view synthesis, yet struggles with complex scenes due to over-reconstruction artifacts.<n>We introduce ReAct-GS, a method founded on the principle of re-activation.<n>We show that ReAct-GS effectively eliminates over-reconstruction artifacts and achieves state-of-the-art performance on standard novel view synthesis metrics.
  arXiv  Detail & Related papers  (2025-10-21T15:38:18Z)
• OracleGS: Grounding Generative Priors for Sparse-View Gaussian Splatting [78.70702961852119]
  OracleGS reconciles generative completeness with regressive fidelity for sparse view Gaussian Splatting.<n>Our approach conditions the powerful generative prior on multi-view geometric evidence, filtering hallucinatory artifacts while preserving plausible completions in under-constrained regions.
  arXiv  Detail & Related papers  (2025-09-27T11:19:32Z)
• RGE-GS: Reward-Guided Expansive Driving Scene Reconstruction via Diffusion Priors [54.81109375939306]
  RGE-GS is a novel expansive reconstruction framework that synergizes diffusion-based generation with reward-guided Gaussian integration.<n>We propose a reward network that learns to identify and prioritize consistently generated patterns prior to reconstruction phases.<n>During the reconstruction process, we devise a differentiated training strategy that automatically adjust Gaussian optimization progress according to scene converge metrics.
  arXiv  Detail & Related papers  (2025-06-28T08:02:54Z)
• DeSiRe-GS: 4D Street Gaussians for Static-Dynamic Decomposition and Surface Reconstruction for Urban Driving Scenes [71.61083731844282]
  We present DeSiRe-GS, a self-supervised gaussian splatting representation.<n>It enables effective static-dynamic decomposition and high-fidelity surface reconstruction in complex driving scenarios.
  arXiv  Detail & Related papers  (2024-11-18T05:49:16Z)

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.