Lyra: Generative 3D Scene Reconstruction via Video Diffusion Model Self-Distillation

• URL: http://arxiv.org/abs/2509.19296v1
• Date: Tue, 23 Sep 2025 17:58:01 GMT
• Title: Lyra: Generative 3D Scene Reconstruction via Video Diffusion Model Self-Distillation
• Authors: Sherwin Bahmani, Tianchang Shen, Jiawei Ren, Jiahui Huang, Yifeng Jiang, Haithem Turki, Andrea Tagliasacchi, David B. Lindell, Zan Gojcic, Sanja Fidler, Huan Ling, Jun Gao, Xuanchi Ren,
• Abstract summary: Current learning-based 3D reconstruction methods rely on the availability of captured real-world multi-view data.<n>We propose a self-distillation framework that distills the implicit 3D knowledge in the video diffusion models into an explicit 3D Gaussian Splatting (3DGS) representation.<n>Our framework achieves state-of-the-art performance in static and dynamic 3D scene generation.
• Score: 87.91642226587294
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: The ability to generate virtual environments is crucial for applications ranging from gaming to physical AI domains such as robotics, autonomous driving, and industrial AI. Current learning-based 3D reconstruction methods rely on the availability of captured real-world multi-view data, which is not always readily available. Recent advancements in video diffusion models have shown remarkable imagination capabilities, yet their 2D nature limits the applications to simulation where a robot needs to navigate and interact with the environment. In this paper, we propose a self-distillation framework that aims to distill the implicit 3D knowledge in the video diffusion models into an explicit 3D Gaussian Splatting (3DGS) representation, eliminating the need for multi-view training data. Specifically, we augment the typical RGB decoder with a 3DGS decoder, which is supervised by the output of the RGB decoder. In this approach, the 3DGS decoder can be purely trained with synthetic data generated by video diffusion models. At inference time, our model can synthesize 3D scenes from either a text prompt or a single image for real-time rendering. Our framework further extends to dynamic 3D scene generation from a monocular input video. Experimental results show that our framework achieves state-of-the-art performance in static and dynamic 3D scene generation.

Related papers

• 3DSPA: A 3D Semantic Point Autoencoder for Evaluating Video Realism [2.6197884751430327]
  We develop an automated evaluation framework for video realism which captures both semantics and coherent 3D structure.<n>Our method, 3DSPA, is 3Dtemporal point autoencoder which integrates 3D point trajectories, depth cues, and DINO semantic features into a unified representation for video evaluation.<n> Experiments show that 3DSPA reliably identifies videos which violate physical laws, is more sensitive to motion artifacts, and aligns more closely with human judgments of video quality and realism.
  arXiv  Detail & Related papers  (2026-02-23T21:00:48Z)
• ManipDreamer3D : Synthesizing Plausible Robotic Manipulation Video with Occupancy-aware 3D Trajectory [56.06314177428745]
  We present ManipDreamer3D for generating plausible 3D-aware robotic manipulation videos from the input image and the text instruction.<n>Our method generates robotic videos with autonomously planned 3D trajectories, significantly reducing human intervention requirements.
  arXiv  Detail & Related papers  (2025-08-29T10:39:06Z)
• Generative Gaussian Splatting: Generating 3D Scenes with Video Diffusion Priors [11.156009461711639]
  Generative Gaussian Splatting (GGS) is a novel approach that integrates a 3D representation with a pre-trained latent video diffusion model.<n>We evaluate our approach on two common benchmark datasets for scene synthesis, RealEstate10K and ScanNet+.
  arXiv  Detail & Related papers  (2025-03-17T15:24:04Z)
• VidBot: Learning Generalizable 3D Actions from In-the-Wild 2D Human Videos for Zero-Shot Robotic Manipulation [53.63540587160549]
  VidBot is a framework enabling zero-shot robotic manipulation using learned 3D affordance from in-the-wild monocular RGB-only human videos.<n> VidBot paves the way for leveraging everyday human videos to make robot learning more scalable.
  arXiv  Detail & Related papers  (2025-03-10T10:04:58Z)
• Dynamic 3D Gaussian Tracking for Graph-Based Neural Dynamics Modeling [10.247075501610492]
  We introduce a framework to learn object dynamics directly from multi-view RGB videos. We train a particle-based dynamics model using Graph Neural Networks. Our method can predict object motions under varying initial configurations and unseen robot actions.
  arXiv  Detail & Related papers  (2024-10-24T17:02:52Z)
• How Far are AI-generated Videos from Simulating the 3D Visual World: A Learned 3D Evaluation Approach [46.85336335756483]
  Learned 3D Evaluation (L3DE) is a method for assessing AI-generated videos' ability to simulate the real world in terms of 3D visual qualities and consistencies.<n>Confidence scores quantify the gap between real and synthetic videos in terms of 3D visual coherence.<n>L3DE extends to broader applications: benchmarking video generation models, serving as a deepfake detector, and enhancing video synthesis by inpainting flagged inconsistencies.
  arXiv  Detail & Related papers  (2024-06-27T23:03:58Z)
• Director3D: Real-world Camera Trajectory and 3D Scene Generation from Text [61.9973218744157]
  We introduce Director3D, a robust open-world text-to-3D generation framework, designed to generate both real-world 3D scenes and adaptive camera trajectories. Experiments demonstrate that Director3D outperforms existing methods, offering superior performance in real-world 3D generation.
  arXiv  Detail & Related papers  (2024-06-25T14:42:51Z)
• DistillNeRF: Perceiving 3D Scenes from Single-Glance Images by Distilling Neural Fields and Foundation Model Features [65.8738034806085]
  DistillNeRF is a self-supervised learning framework for understanding 3D environments in autonomous driving scenes. Our method is a generalizable feedforward model that predicts a rich neural scene representation from sparse, single-frame multi-view camera inputs.
  arXiv  Detail & Related papers  (2024-06-17T21:15:13Z)
• OneTo3D: One Image to Re-editable Dynamic 3D Model and Video Generation [0.0]
  One image to editable dynamic 3D model and video generation is novel direction and change in the research area of single image to 3D representation or 3D reconstruction of image. We propose the OneTo3D, a method and theory to used one single image to generate the editable 3D model and generate the targeted semantic continuous time-unlimited 3D video.
  arXiv  Detail & Related papers  (2024-05-10T15:44:11Z)
• Learning 3D Particle-based Simulators from RGB-D Videos [15.683877597215494]
  We propose a method for learning simulators directly from observations. Visual Particle Dynamics (VPD) jointly learns a latent particle-based representation of 3D scenes. Unlike existing 2D video prediction models, VPD's 3D structure enables scene editing and long-term predictions.
  arXiv  Detail & Related papers  (2023-12-08T20:45:34Z)
• PonderV2: Pave the Way for 3D Foundation Model with A Universal Pre-training Paradigm [111.16358607889609]
  We introduce a novel universal 3D pre-training framework designed to facilitate the acquisition of efficient 3D representation.<n>For the first time, PonderV2 achieves state-of-the-art performance on 11 indoor and outdoor benchmarks, implying its effectiveness.
  arXiv  Detail & Related papers  (2023-10-12T17:59:57Z)
• T3VIP: Transformation-based 3D Video Prediction [49.178585201673364]
  We propose a 3D video prediction (T3VIP) approach that explicitly models the 3D motion by decomposing a scene into its object parts. Our model is fully unsupervised, captures the nature of the real world, and the observational cues in image and point cloud domains constitute its learning signals. To the best of our knowledge, our model is the first generative model that provides an RGB-D video prediction of the future for a static camera.
  arXiv  Detail & Related papers  (2022-09-19T15:01:09Z)

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.