mlsanityData to 14 Oct 2026

DeFlow: Decoder of Scene Flow Network in Autonomous Driving

arXiv 2401.16122 ↗First posted 29 Jan 2024cs.CV cs.ROcited by 30

Not on the map. Why?

Abstract (arXiv)

Scene flow estimation determines a scene's 3D motion field, by predicting the motion of points in the scene, especially for aiding tasks in autonomous driving. Many networks with large-scale point clouds as input use voxelization to create a pseudo-image for real-time running. However, the voxelization process often results in the loss of point-specific features. This gives rise to a challenge in recovering those features for scene flow tasks. Our paper introduces DeFlow which enables a transition from voxel-based features to point features using Gated Recurrent Unit (GRU) refinement. To further enhance scene flow estimation performance, we formulate a novel loss function that accounts for the data imbalance between static and dynamic points. Evaluations on the Argoverse 2 scene flow task reveal that DeFlow achieves state-of-the-art results on large-scale point cloud data, demonstrating that our network has better performance and efficiency compared to others. The code is open-sourced at https://github.com/KTH-RPL/deflow.

Key facts. Cited by 30 papers in the map (data through 14 Oct 2026). Builds on 28 of its 36 references that are in the map. First posted on arXiv 29 Jan 2024.

Builds on

28 of 36 references.

References · 28

  1. 2016PointNet: Deep Learning on Point Sets for 3D Classification and Segmentationcited by 5,248
  2. 2015ShapeNet: An Information-Rich 3D Model Repositorycited by 2,897
  3. 2019Scalability in Perception for Autonomous Driving: Waymo Open Datasetcited by 2,451
  4. 2020RAFT: Recurrent All-Pairs Field Transforms for Optical Flowcited by 2,415
  5. 2018PointPillars: Fast Encoders for Object Detection from Point Cloudscited by 1,715
  6. 2015Multi-Scale Context Aggregation by Dilated Convolutionscited by 1,133
  7. 2015FlowNet: Learning Optical Flow with Convolutional Networkscited by 1,059
  8. 2017PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volumecited by 968
  9. 2015A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimationcited by 881
  10. 2023Argoverse 2: Next Generation Datasets for Self-Driving Perception and Forecastingcited by 707
  11. 2021GMFlow: Learning Optical Flow via Global Matchingcited by 327
  12. 2019End-to-End Multi-View Fusion for 3D Object Detection in LiDAR Point Cloudscited by 171
16 more
  1. 2019Iterative Residual Refinement for Joint Optical Flow and Occlusion Estimationcited by 131
  2. 2022Point-to-Voxel Knowledge Distillation for LiDAR Semantic Segmentationcited by 119
  3. 2021Neural Scene Flow Priorcited by 91
  4. 2022CRAFT: Cross-Attentional Flow Transformer for Robust Optical Flowcited by 82
  5. 2020PV-RAFT: Point-Voxel Correlation Fields for Scene Flow Estimation of Point Cloudscited by 74
  6. 2022Motion Inspired Unsupervised Perception and Prediction in Autonomous Drivingcited by 59
  7. 2021Scalable Scene Flow from Point Clouds in the Real Worldcited by 50
  8. 2022Patchwork++: Fast and Robust Ground Segmentation Solving Partial Under-Segmentation Using 3D Point Cloudcited by 47
  9. 2021Self-Supervised Pillar Motion Learning for Autonomous Drivingcited by 44
  10. 2023Fast Neural Scene Flowcited by 41
  11. 2022Bi-PointFlowNet: Bidirectional Learning for Point Cloud Based Scene Flow Estimationcited by 38
  12. 2022Open-world Semantic Segmentation for LIDAR Point Cloudscited by 29
  13. 2022SCOOP: Self-Supervised Correspondence and Optimization-Based Scene Flowcited by 28
  14. 2023Dynablox: Real-time Detection of Diverse Dynamic Objects in Complex Environmentscited by 12
  15. 2023A Dynamic Points Removal Benchmark in Point Cloud Mapscited by 8
  16. 2022GMA3D: Local-Global Attention Learning to Estimate Occluded Motions of Scene Flowcited by 3

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 6; 2025: 14; 2026: 10.

30 papers.

Sort:

By area:

  1. 2024SeFlow: A Self-Supervised Scene Flow Method in Autonomous Drivingcited by 39
  2. 2024Flow4D: Leveraging 4D Voxel Network for LiDAR Scene Flow Estimationcited by 23
  3. 2024I Can't Believe It's Not Scene Flow!cited by 22
  4. 2024Neural Eulerian Scene Flow Fieldscited by 15
  5. 2025SSF: Sparse Long-Range Scene Flow for Autonomous Drivingcited by 13
  6. 2025HiMo: High-Speed Objects Motion Compensation in Point Cloudscited by 12
  7. 2024DUFOMap: Efficient Dynamic Awareness Mappingcited by 11
  8. 2025VoteFlow: Enforcing Local Rigidity in Self-Supervised Scene Flowcited by 11
  9. 2025Floxels: Fast Unsupervised Voxel Based Scene Flow Estimationcited by 10
  10. 2025AGO: Adaptive Grounding for Open World 3D Occupancy Predictioncited by 9
  11. 2025MambaFlow: A Novel and Flow-guided State Space Model for Scene Flow Estimationcited by 8
  12. 2025DeltaFlow: An Efficient Multi-frame Scene Flow Estimation Methodcited by 8
18 more
  1. 2024BeautyMap: Binary-Encoded Adaptable Ground Matrix for Dynamic Points Removal in Global Mapscited by 6
  2. 2026TeFlow: Enabling Multi-frame Supervision for Self-Supervised Feed-forward Scene Flow Estimationcited by 6
  3. 2025DoGFlow: Self-Supervised LiDAR Scene Flow via Cross-Modal Doppler Guidancecited by 4
  4. 2025TARS: Traffic-Aware Radar Scene Flow Estimationcited by 3
  5. 2025SemanticFlow: A Self-Supervised Framework for Joint Scene Flow Prediction and Instance Segmentation in Dynamic Environmentscited by 3
  6. 2025RaLiFlow: Scene Flow Estimation with 4D Radar and LiDAR Point Cloudscited by 3
  7. 2026RegFormer++: An Efficient Large-Scale 3D LiDAR Point Registration Network with Projection-Aware 2D Transformercited by 3
  8. 2025SSF-PAN: Semantic Scene Flow-Based Perception for Autonomous Navigation in Traffic Scenarioscited by 2
  9. 2026AsyncBEV: Cross-modal Flow Alignment in Asynchronous 3D Object Detectioncited by 2
  10. 2026SynFlow: Scaling Up LiDAR Scene Flow Estimation with Synthetic Datacited by 2
  11. 2025Temporal Overlapping Prediction: A Self-supervised Pre-training Method for LiDAR Moving Object Segmentationcited by 0
  12. 2025UniFlow: Zero-Shot LiDAR Scene Flow for Autonomous Vehiclescited by 0
  13. 2026Motion Cues from Image-based Point Tracking for LiDAR Scene Flow Estimationcited by 0
  14. 2026Weakly Supervised Cross-Modal Learning for 4D Radar Scene Flow Estimationcited by 0
  15. 2026CorrelationFlow: A Training-Free Geometric Approach for LiDAR Scene Flow Estimationcited by 0
  16. 2026RVLoss: Runoff Vote Loss for Self-Supervised LiDAR Scene Flow Estimationcited by 0
  17. 2026AevaScenes: An FMCW LiDAR Dataset and Benchmark for Long-Range Perceptioncited by 0
  18. 2026MeshBind: Point-to-Mesh Binding for Non-Rigid Scene Flow Automatic Annotationcited by 0