mlsanityData to 14 Oct 2026

What Matters for 3D Scene Flow Network

arXiv 2207.09143 ↗First posted 19 Jul 2022cs.CVcited by 31

Not on the map. Why?

Abstract (arXiv)

3D scene flow estimation from point clouds is a low-level 3D motion perception task in computer vision. Flow embedding is a commonly used technique in scene flow estimation, and it encodes the point motion between two consecutive frames. Thus, it is critical for the flow embeddings to capture the correct overall direction of the motion. However, previous works only search locally to determine a soft correspondence, ignoring the distant points that turn out to be the actual matching ones. In addition, the estimated correspondence is usually from the forward direction of the adjacent point clouds, and may not be consistent with the estimated correspondence acquired from the backward direction. To tackle these problems, we propose a novel all-to-all flow embedding layer with backward reliability validation during the initial scene flow estimation. Besides, we investigate and compare several design choices in key components of the 3D scene flow network, including the point similarity calculation, input elements of predictor, and predictor & refinement level design. After carefully choosing the most effective designs, we are able to present a model that achieves the state-of-the-art performance on FlyingThings3D and KITTI Scene Flow datasets. Our proposed model surpasses all existing methods by at least 38.2% on FlyingThings3D dataset and 24.7% on KITTI Scene Flow dataset for EPE3D metric. We release our codes at https://github.com/IRMVLab/3DFlow.

Key facts. Cited by 31 papers in the map (data through 14 Oct 2026). Builds on 33 of its 54 references that are in the map. First posted on arXiv 19 Jul 2022.

Builds on

33 of 54 references.

References · 33

  1. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  2. 2016PointNet: Deep Learning on Point Sets for 3D Classification and Segmentationcited by 5,248
  3. 2017PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Spacecited by 4,401
  4. 2020RAFT: Recurrent All-Pairs Field Transforms for Optical Flowcited by 2,415
  5. 2017PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volumecited by 968
  6. 2015A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimationcited by 881
  7. 2018FlowNet3D: Learning Scene Flow in 3D Point Cloudscited by 242
  8. 2018SPLATNet: Sparse Lattice Networks for Point Cloud Processingcited by 158
  9. 2020FLOT: Scene Flow on Point Clouds Guided by Optimal Transportcited by 129
  10. 2020What Matters in Unsupervised Optical Flowcited by 124
  11. 2019MeteorNet: Deep Learning on Dynamic 3D Point Cloud Sequencescited by 117
  12. 2019HPLFlowNet: Hierarchical Permutohedral Lattice FlowNet for Scene Flow Estimation on Large-scale Point Cloudscited by 109
21 more
  1. 2019Just Go with the Flow: Self-Supervised Scene Flow Estimationcited by 98
  2. 2020PV-RAFT: Point-Voxel Correlation Fields for Scene Flow Estimation of Point Cloudscited by 74
  3. 2020FlowStep3D: Model Unrolling for Self-Supervised Scene Flow Estimationcited by 71
  4. 2018Occlusions, Motion and Depth Boundaries with a Generic Network for Disparity, Optical Flow or Scene Flow Estimationcited by 69
  5. 2020SOE-Net: A Self-Attention and Orientation Encoding Network for Point Cloud based Place Recognitioncited by 66
  6. 2020Self-Supervised Monocular Scene Flow Estimationcited by 64
  7. 2019Deep Rigid Instance Scene Flowcited by 54
  8. 2021Learning to Segment Rigid Motions from Two Framescited by 46
  9. 2021Self-Point-Flow: Self-Supervised Scene Flow Estimation from Point Clouds with Optimal Transport and Random Walkcited by 39
  10. 2021HCRF-Flow: Scene Flow from Point Clouds with Continuous High-order CRFs and Position-aware Flow Embeddingcited by 36
  11. 2020Hierarchical Attention Learning of Scene Flow in 3D Point Cloudscited by 35
  12. 2020PWCLO-Net: Deep LiDAR Odometry in 3D Point Clouds Using Hierarchical Embedding Mask Optimizationcited by 35
  13. 2021FESTA: Flow Estimation via Spatial-Temporal Attention for Scene Point Cloudscited by 30
  14. 2021Self-Supervised Multi-Frame Monocular Scene Flowcited by 29
  15. 2021Efficient 3D Deep LiDAR Odometrycited by 25
  16. 2020PointTrackNet: An End-to-End Network For 3-D Object Detection and Tracking From Point Cloudscited by 20
  17. 2020Anchor-Based Spatio-Temporal Attention 3D Convolutional Networks for Dynamic 3D Point Cloud Sequencescited by 9
  18. 2022Unsupervised Learning of 3D Scene Flow from Monocular Cameracited by 7
  19. 2020Spherical Interpolated Convolutional Network with Distance-Feature Density for 3D Semantic Segmentation of Point Cloudscited by 6
  20. 2014Permutohedral Lattice CNNscited by 4
  21. 2022Interactive Multi-scale Fusion of 2D and 3D Features for Multi-object Trackingcited by 4

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2022: 2; 2023: 13; 2024: 11; 2025: 2; 2026: 3.

31 papers.

Sort:

By area:

  1. 2023RegFormer: An Efficient Projection-Aware Transformer Network for Large-Scale Point Cloud Registrationcited by 37
  2. 2023Hidden Gems: 4D Radar Scene Flow Learning Using Cross-Modal Supervisioncited by 36
  3. 2022SCOOP: Self-Supervised Correspondence and Optimization-Based Scene Flowcited by 28
  4. 2023Self-Supervised 3D Scene Flow Estimation Guided by Superpointscited by 24
  5. 2025Zero-Shot Monocular Scene Flow Estimation in the Wildcited by 22
  6. 2026Align then Adapt: Rethinking Parameter-Efficient Transfer Learning in 4D Perceptioncited by 18
  7. 2023Re-Evaluating LiDAR Scene Flow for Autonomous Drivingcited by 17
  8. 20243DSFLabelling: Boosting 3D Scene Flow Estimation by Pseudo Auto-labellingcited by 16
  9. 2023GMSF: Global Matching Scene Flowcited by 14
  10. 2023DELFlow: Dense Efficient Learning of Scene Flow for Large-Scale Point Cloudscited by 14
  11. 2023Multi-Body Neural Scene Flowcited by 12
  12. 2023DifFlow3D: Toward Robust Uncertainty-Aware Scene Flow Estimation with Diffusion Modelcited by 11
19 more
  1. 2024ManiFoundation Model for General-Purpose Robotic Manipulation of Contact Synthesis with Arbitrary Objects and Robotscited by 11
  2. 2023End-to-end 2D-3D Registration between Image and LiDAR Point Cloud for Vehicle Localizationcited by 9
  3. 2024DiffSF: Diffusion Models for Scene Flow Estimationcited by 6
  4. 2024OptFlow: Fast Optimization-based Scene Flow Estimation without Supervisioncited by 5
  5. 2024FlowMamba: Learning Point Cloud Scene Flow with Global Motion Propagationcited by 5
  6. 2023EMR-MSF: Self-Supervised Recurrent Monocular Scene Flow Exploiting Ego-Motion Rigiditycited by 4
  7. 2024Self-Supervised Multi-Frame Neural Scene Flowcited by 3
  8. 2026RegFormer++: An Efficient Large-Scale 3D LiDAR Point Registration Network with Projection-Aware 2D Transformercited by 3
  9. 2022Unsupervised Learning of 3D Scene Flow with 3D Odometry Assistancecited by 2
  10. 2023Self-Supervised 3D Scene Flow Estimation and Motion Prediction using Local Rigidity Priorcited by 2
  11. 2025Weakly and Self-Supervised Class-Agnostic Motion Prediction for Autonomous Drivingcited by 2
  12. 2023Exploiting Implicit Rigidity Constraints via Weight-Sharing Aggregation for Scene Flow Estimation from Point Cloudscited by 1
  13. 2023SSFlowNet: Semi-supervised Scene Flow Estimation On Point Clouds With Pseudo Labelcited by 1
  14. 2024CMU-Flownet: Exploring Point Cloud Scene Flow Estimation in Occluded Scenariocited by 1
  15. 2024RMS-FlowNet++: Efficient and Robust Multi-Scale Scene Flow Estimation for Large-Scale Point Cloudscited by 1
  16. 2024STARFlow: Spatial Temporal Feature Re-embedding with Attentive Learning for Real-world Scene Flowcited by 0
  17. 2024EgoFlowNet: Non-Rigid Scene Flow from Point Clouds with Ego-Motion Supportcited by 0
  18. 2024SSRFlow: Semantic-aware Fusion with Spatial Temporal Re-embedding for Real-world Scene Flowcited by 0
  19. 2026PCSTracker: Long-Term Scene Flow Estimation for Point Cloud Sequencescited by 0