mlsanityData to 7 Oct 2026

LiFlow: Flow Matching for 3D LiDAR Scene Completion

arXiv 2602.02232 ↗First posted 2 Feb 2026cs.CVcited by 2

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

Abstract (arXiv)

In autonomous driving scenarios, the collected LiDAR point clouds can be challenged by occlusion and long-range sparsity, limiting the perception of autonomous driving systems. Scene completion methods can infer the missing parts of incomplete 3D LiDAR scenes. Recent methods adopt local point-level denoising diffusion probabilistic models, which require predicting Gaussian noise, leading to a mismatch between training and inference initial distributions. This paper introduces the first flow matching framework for 3D LiDAR scene completion, improving upon diffusion-based methods by ensuring consistent initial distributions between training and inference. The model employs a nearest neighbor flow matching loss and a Chamfer distance loss to enhance both local structure and global coverage in the alignment of point clouds. LiFlow achieves state-of-the-art performance across multiple metrics. Code: https://github.com/matteandre/LiFlow.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in 3D, point cloud and scene generation, one of 14 directions in Diffusion models. Builds on 31 of its 36 references that are in the map. First posted on arXiv 2 Feb 2026.

Builds on

31 of 36 references.

Closest ancestors · 7

Same direction, newest first

  1. 2025LiDPM: Rethinking Point Diffusion for Lidar Scene Completion24 Apr
  2. 2025Efficient Point Clouds Upsampling via Flow Matching25 Jan
  3. 2024GaussianAnything: Interactive Point Cloud Flow Matching For 3D Object Generation12 Nov
  4. 2024Towards Realistic Scene Generation with LiDAR Diffusion Models31 Mar
  5. 2024Scaling Diffusion Models to Real-World 3D LiDAR Scene Completion20 Mar
  6. 2023Diffusion Probabilistic Models for Scene-Scale 3D Categorical Data2 Jan
  7. 2021Diffusion Probabilistic Models for 3D Point Cloud Generation2 Mar

Other references · 24

  1. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  2. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  3. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 8,964
  4. 2015Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shiftcited by 6,979
  5. 2022Classifier-Free Diffusion Guidancecited by 5,717
  6. 2022Flow Matching for Generative Modelingcited by 5,426
  7. 2022DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Stepscited by 1,961
  8. 20194D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networkscited by 1,201
  9. 2016Instance Normalization: The Missing Ingredient for Fast Stylizationcited by 1,187
  10. 2023Improving and generalizing flow-based generative models with minibatch optimal transportcited by 835
  11. 20213D Shape Generation and Completion through Point-Voxel Diffusioncited by 516
  12. 2016Semantic Scene Completion from a Single Depth Imagecited by 468
12 more
  1. 2020Sparse Single Sweep LiDAR Point Cloud Segmentation via Learning Contextual Shape Priors from Scene Completioncited by 222
  2. 2017Sparse-to-Dense: Depth Prediction from Sparse Depth Samples and a Single Imagecited by 159
  3. 2018Self-supervised Sparse-to-Dense: Self-supervised Depth Completion from LiDAR and Monocular Cameracited by 140
  4. 2023SCPNet: Semantic Scene Completion on Point Cloudcited by 80
  5. 2018Sparse and Dense Data with CNNs: Depth Completion and Semantic Segmentationcited by 76
  6. 2023UltraLiDAR: Learning Compact Representations for LiDAR Completion and Generationcited by 76
  7. 2023Boosting Latent Diffusion with Flow Matchingcited by 44
  8. 2023LODE: Locally Conditioned Eikonal Implicit Scene Completion from Sparse LiDARcited by 31
  9. 2022Automatic Labeling to Generate Training Data for Online LiDAR-based Moving Object Segmentationcited by 27
  10. 2025Diff2Flow: Training Flow Matching Models via Diffusion Model Alignmentcited by 27
  11. 2020Depth Completion via Inductive Fusion of Planar LIDAR and Monocular Cameracited by 22
  12. 2023Building Volumetric Beliefs for Dynamic Environments Exploiting Map-Based Moving Object Segmentationcited by 15

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026FPSGen: Flexible Point Cloud Scene Generation with BEV-Supported Transport Flowscited by 0
  2. 2026Towards Real-Time and Adaptable LiDAR Scene Completioncited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024DiffSSC: Semantic LiDAR Scan Completion using Denoising Diffusion Probabilistic Models
  2. 2026CloudDiffusion: Diffusion-Based Scene Completion in the Point Cloud Domain
  3. 2024Distilling Diffusion Models to Efficient 3D LiDAR Scene Completion
  4. 2025LiNeXt: Revisiting LiDAR Completion with Efficient Non-Diffusion Architectures
  5. 2025Diffusion Distillation With Direct Preference Optimization For Efficient 3D LiDAR Scene Completion
  6. 2024Fast LiDAR Upsampling using Conditional Diffusion Models
  7. 2023LiDAR Data Synthesis with Denoising Diffusion Probabilistic Models
  8. 2026PatchScene: Patch-based Voxel Diffusion for Large-Scale Scene Completion