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

• URL: http://arxiv.org/abs/2508.19204v1
• Date: Tue, 26 Aug 2025 17:04:49 GMT
• Title: LSD-3D: Large-Scale 3D Driving Scene Generation with Geometry Grounding
• Abstract summary: We present a method that directly generates large-scale 3D driving scenes with accurate geometry.<n>The proposed method combines the generation of a proxy geometry and environment representation with score distillation from learned 2D image priors.<n>We find that this approach allows for high controllability, enabling the prompt-guided geometry and high-fidelity texture and structure.
• Score: 34.74478301165912
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large-scale scene data is essential for training and testing in robot learning. Neural reconstruction methods have promised the capability of reconstructing large physically-grounded outdoor scenes from captured sensor data. However, these methods have baked-in static environments and only allow for limited scene control -- they are functionally constrained in scene and trajectory diversity by the captures from which they are reconstructed. In contrast, generating driving data with recent image or video diffusion models offers control, however, at the cost of geometry grounding and causality. In this work, we aim to bridge this gap and present a method that directly generates large-scale 3D driving scenes with accurate geometry, allowing for causal novel view synthesis with object permanence and explicit 3D geometry estimation. The proposed method combines the generation of a proxy geometry and environment representation with score distillation from learned 2D image priors. We find that this approach allows for high controllability, enabling the prompt-guided geometry and high-fidelity texture and structure that can be conditioned on map layouts -- producing realistic and geometrically consistent 3D generations of complex driving scenes.

Related papers

• ScaleVid: Geometry-Aware Video Object Scaling with Mesh-Free Inference [13.10996719828907]
  Geometry-aware video object scaling aims to anisotropically resize the object along object-centric axes while preserving geometric plausibility, temporal coherence, and background consistency.<n>We present a two-stage training framework that decouples geometry-aware foreground transformation from background preservation and realistic video composition.<n>In both stages, geometrically perturbed pseudo-sources are constructed from real videos, while the original complete videos are retained as reconstruction targets.
  arXiv  Detail & Related papers  (2026-08-12T16:28:26Z)
• Lyra 2.0: Explorable Generative 3D Worlds [77.45279013687427]
  We present Lyra 2.0, a framework for generating persistent, explorable 3D worlds at scale.<n>To address spatial forgetting, we maintain per-frame 3D geometry and use it solely for information routing.<n>We train with self-augmented histories that expose the model to its own degraded outputs, teaching it to correct drift rather than propagate it.
  arXiv  Detail & Related papers  (2026-04-14T17:59:44Z)
• GeoWorld: Unlocking the Potential of Geometry Models to Facilitate High-Fidelity 3D Scene Generation [68.02988074681427]
  Previous works leveraging video models for image-to-3D scene generation tend to suffer from geometric distortions and blurry content.<n>In this paper, we renovate the pipeline of image-to-3D scene generation by unlocking the potential of geometry models.<n>Our GeoWorld can generate high-fidelity 3D scenes from a single image and a given camera trajectory, outperforming prior methods both qualitatively and quantitatively.
  arXiv  Detail & Related papers  (2025-11-28T13:55:45Z)
• Constructing a 3D Town from a Single Image [23.231661811526955]
  3DTown is a training-free framework designed to synthesize realistic and coherent 3D scenes from a single top-down view.<n>We decompose the input image into overlapping regions and generate each using a pretrained 3D object generator.<n>Our results demonstrate that high-quality 3D town generation is achievable from a single image using a principled, training-free approach.
  arXiv  Detail & Related papers  (2025-05-21T17:10:47Z)
• Incorporating dense metric depth into neural 3D representations for view synthesis and relighting [25.028859317188395]
  In robotic applications, dense metric depth can often be measured directly using stereo and illumination can be controlled. In this work we demonstrate a method to incorporate dense metric depth into the training of neural 3D representations. We also discuss a multi-flash stereo camera system developed to capture the necessary data for our pipeline and show results on relighting and view synthesis.
  arXiv  Detail & Related papers  (2024-09-04T20:21:13Z)
• Outdoor Scene Extrapolation with Hierarchical Generative Cellular Automata [70.9375320609781]
  We aim to generate fine-grained 3D geometry from large-scale sparse LiDAR scans, abundantly captured by autonomous vehicles (AV) We propose hierarchical Generative Cellular Automata (hGCA), a spatially scalable 3D generative model, which grows geometry with local kernels following, in a coarse-to-fine manner, equipped with a light-weight planner to induce global consistency.
  arXiv  Detail & Related papers  (2024-06-12T14:56:56Z)
• Invisible Stitch: Generating Smooth 3D Scenes with Depth Inpainting [75.7154104065613]
  We introduce a novel depth completion model, trained via teacher distillation and self-training to learn the 3D fusion process. We also introduce a new benchmarking scheme for scene generation methods that is based on ground truth geometry.
  arXiv  Detail & Related papers  (2024-04-30T17:59:40Z)
• MagicDrive: Street View Generation with Diverse 3D Geometry Control [82.69871576797166]
  We introduce MagicDrive, a novel street view generation framework, offering diverse 3D geometry controls. Our design incorporates a cross-view attention module, ensuring consistency across multiple camera views.
  arXiv  Detail & Related papers  (2023-10-04T06:14:06Z)
• FrozenRecon: Pose-free 3D Scene Reconstruction with Frozen Depth Models [67.96827539201071]
  We propose a novel test-time optimization approach for 3D scene reconstruction. Our method achieves state-of-the-art cross-dataset reconstruction on five zero-shot testing datasets.
  arXiv  Detail & Related papers  (2023-08-10T17:55:02Z)
• Neural 3D Scene Reconstruction from Multiple 2D Images without 3D Supervision [41.20504333318276]
  We propose a novel neural reconstruction method that reconstructs scenes using sparse depth under the plane constraints without 3D supervision. We introduce a signed distance function field, a color field, and a probability field to represent a scene. We optimize these fields to reconstruct the scene by using differentiable ray marching with accessible 2D images as supervision.
  arXiv  Detail & Related papers  (2023-06-30T13:30:48Z)
• 3D Data Augmentation for Driving Scenes on Camera [50.41413053812315]
  We propose a 3D data augmentation approach termed Drive-3DAug, aiming at augmenting the driving scenes on camera in the 3D space. We first utilize Neural Radiance Field (NeRF) to reconstruct the 3D models of background and foreground objects. Then, augmented driving scenes can be obtained by placing the 3D objects with adapted location and orientation at the pre-defined valid region of backgrounds.
  arXiv  Detail & Related papers  (2023-03-18T05:51:05Z)

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.