mlsanityData to 14 Oct 2026

Generative Semantic Scene Completion

arXiv 2608.26737 ↗First posted 27 Aug 2026cs.CV cs.LG cs.ROcited by 0

On the map: Diffusion models → 3D, point cloud and scene generation

Abstract (arXiv)

Outdoor LiDAR semantic scene completion (SSC) recovers a dense semantic voxel grid from a scan observing 1% of the target volume, under class imbalance beyond 7,000x. We recast SSC as generative semantic scene completion (GSSC): a single discrete-diffusion formulation in three roles. First, paired sparse-dense scene synthesis (PS3) generates matched sparse LiDAR observations with their dense semantic completions, addressing the long tail at its source and yielding the PS3-SemanticKITTI corpus we train on alongside SemanticKITTI. Second, semantic-guided generative scene completion (SGSC) generates the scene from noise with multinomial discrete diffusion, conditioned on the sparse scan through a bird's-eye-view semantic map and a sparse 3D feature stream. Third, the same framework instead refines an existing completion in one flow-matching step: structured source discrete diffusion (S2D2). S2D2 improves the mIoU of SGSC's own output and every external SSC base tested, without base retraining or test-time adaptation. On the strongest base, one step without test-time augmentation reaches 38.8% mIoU on the SemanticKITTI hidden test. To our knowledge that is the best causal, single-sweep, single-sample result on that leaderboard, +2.1 pp over the previous best published score under the same restriction. Four correction steps with eight-view test-time augmentation reach 39.2%, outside that restriction.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in 3D, point cloud and scene generation, one of 14 directions in Diffusion models. Builds on 56 of its 72 references that are in the map. First posted on arXiv 27 Aug 2026.

Builds on

56 of 72 references.

Closest ancestors · 10

Same direction, newest first

  1. 2024Distilling Diffusion Models to Efficient 3D LiDAR Scene Completion4 Dec
  2. 2024DiffSSC: Semantic LiDAR Scan Completion using Denoising Diffusion Probabilistic Models26 Sep
  3. 2024Towards Realistic Scene Generation with LiDAR Diffusion Models31 Mar
  4. 2024Scaling Diffusion Models to Real-World 3D LiDAR Scene Completion20 Mar
  5. 2024SemCity: Semantic Scene Generation with Triplane Diffusion12 Mar
  6. 2023Pyramid Diffusion for Fine 3D Large Scene Generation20 Nov
  7. 2023DiffuScene: Denoising Diffusion Models for Generative Indoor Scene Synthesis24 Mar
  8. 2023Diffusion Probabilistic Models for Scene-Scale 3D Categorical Data2 Jan
  9. 2022LION: Latent Point Diffusion Models for 3D Shape Generation12 Oct
  10. 2021Diffusion Probabilistic Models for 3D Point Cloud Generation2 Mar

Other references · 46

  1. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  2. 2020Denoising Diffusion Implicit Modelscited by 9,781
  3. 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,255
  4. 2022Flow Matching for Generative Modelingcited by 5,702
  5. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flowcited by 3,700
  6. 2017CARLA: An Open Urban Driving Simulatorcited by 2,449
  7. 2017FiLM: Visual Reasoning with a General Conditioning Layercited by 2,333
  8. 2023Consistency Modelscited by 1,822
  9. 2021Structured Denoising Diffusion Models in Discrete State-Spacescited by 1,719
  10. 2019Class-Balanced Loss Based on Effective Number of Samplescited by 1,245
  11. 2022Planning-oriented Autonomous Drivingcited by 1,087
  12. 2023VAD: Vectorized Scene Representation for Efficient Autonomous Drivingcited by 622
34 more
  1. 2021Argmax Flows and Multinomial Diffusion: Learning Categorical Distributionscited by 574
  2. 20213D Shape Generation and Completion through Point-Voxel Diffusioncited by 516
  3. 2016Semantic Scene Completion from a Single Depth Imagecited by 472
  4. 2023ResShift: Efficient Diffusion Model for Image Super-resolution by Residual Shiftingcited by 365
  5. 2024Discrete Flow Matchingcited by 317
  6. 2024Generative Flows on Discrete State-Spaces: Enabling Multimodal Flows with Applications to Protein Co-Designcited by 305
  7. 2022Cold Diffusion: Inverting Arbitrary Image Transforms Without Noisecited by 276
  8. 2022BBDM: Image-to-image Translation with Brownian Bridge Diffusion Modelscited by 225
  9. 2020Sparse Single Sweep LiDAR Point Cloud Segmentation via Learning Contextual Shape Priors from Scene Completioncited by 223
  10. 2023PDE-Refiner: Achieving Accurate Long Rollouts with Neural PDE Solverscited by 216
  11. 2020LMSCNet: Lightweight Multiscale 3D Semantic Completioncited by 188
  12. 2023Denoising Diffusion Bridge Modelscited by 178
  13. 2023Inversion by Direct Iteration: An Alternative to Denoising Diffusion for Image Restorationcited by 165
  14. 2020LiDARsim: Realistic LiDAR Simulation by Leveraging the Real Worldcited by 137
  15. 2020SemanticPOSS: A Point Cloud Dataset with Large Quantity of Dynamic Instancescited by 123
  16. 2020S3CNet: A Sparse Semantic Scene Completion Network for LiDAR Point Cloudscited by 102
  17. 2023SCPNet: Semantic Scene Completion on Point Cloudcited by 80
  18. 2023UltraLiDAR: Learning Compact Representations for LiDAR Completion and Generationcited by 78
  19. 2023SSCBench: A Large-Scale 3D Semantic Scene Completion Benchmark for Autonomous Drivingcited by 75
  20. 2020Semantic Scene Completion using Local Deep Implicit Functions on LiDAR Datacited by 74
  21. 20213D Semantic Scene Completion: a Surveycited by 74
  22. 2022PolarMix: A General Data Augmentation Technique for LiDAR Point Cloudscited by 74
  23. 2024LidarDM: Generative LiDAR Simulation in a Generated Worldcited by 54
  24. 2023PaSCo: Urban 3D Panoptic Scene Completion with Uncertainty Awarenesscited by 42
  25. 2022MotionSC: Data Set and Network for Real-Time Semantic Mapping in Dynamic Environmentscited by 40
  26. 2021Semantic Segmentation-assisted Scene Completion for LiDAR Point Cloudscited by 37
  27. 2023SSC-RS: Elevate LiDAR Semantic Scene Completion with Representation Separation and BEV Fusioncited by 30
  28. 2023SegRefiner: Towards Model-Agnostic Segmentation Refinement with Discrete Diffusion Processcited by 29
  29. 2024LSK3DNet: Towards Effective and Efficient 3D Perception with Large Sparse Kernelscited by 15
  30. 2024Offboard Occupancy Refinement with Hybrid Propagation for Autonomous Drivingcited by 12
  31. 2024TALoS: Enhancing Semantic Scene Completion via Test-time Adaptation on the Line of Sightcited by 12
  32. 2025Enhancing 3D Semantic Scene Completion with a Refinement Modulecited by 3
  33. 2026Exploring Easy Boosts for Lidar Semantic Scene Completioncited by 2
  34. 2025Octree Diffusion for Semantic Scene Generation and Completioncited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025LiDPM: Rethinking Point Diffusion for Lidar Scene Completion
  2. 2026PatchScene: Patch-based Voxel Diffusion for Large-Scale Scene Completion
  3. 2025SPIRAL: Semantic-Aware Progressive LiDAR Scene Generation and Understanding
  4. 2026Iterate or Widen? When Test-Time Refinement Helps LiDAR Scene Completion: A Controlled Study of Evidence Geometry, Training Coverage, and Compute
  5. 2026FlowSSC: Universal Generative Monocular Semantic Scene Completion via One-Step Latent Diffusion
  6. 2025LiNeXt: Revisiting LiDAR Completion with Efficient Non-Diffusion Architectures
  7. 2026CloudDiffusion: Diffusion-Based Scene Completion in the Point Cloud Domain
  8. 2024Label-efficient Semantic Scene Completion with Scribble Annotations