Research map / Point clouds / Point cloud classification
Scene flow and dynamic point clouds
Estimate motion and process sequences of dynamic point clouds.
89 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2019–2026) · an approach in Point cloud classification, keyword cluster “motion · estimation · scene”
State History Benchmarks Tried Baselines Checklist Preprints Who Export
Where it stands
Few recent reports The tracked benchmarks have at most one report from 2025–26 that could beat the old record, too few to say whether they are still moving.
29 papers in 2023–24 and 26 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 12.5% to 10.0%.
No longer reported by papers here: KITTI_o (EPE, last in 2023); KITTI_t (EPE, last in 2023).
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2019 · FlowNet3D: Learning Scene Flow in 3D Point Clouds (CVPR 2019 · 526 citations)
- 2020 · P2B: Point-to-Box Network for 3D Object Tracking in Point Clouds (CVPR 2020 · 173 citations)
- 2021 · Point 4D Transformer Networks for Spatio-Temporal Modeling in Point Cloud Videos (CVPR 2021 · 210 citations)
- 2022 · Point-NeRF: Point-Based Neural Radiance Fields (CVPR 2022 · 561 citations)
- 2023 · Efficient Hierarchical Entropy Model for Learned Point Cloud Compression (CVPR 2023 · 67 citations)
- 2024 · Visual Point Cloud Forecasting enables Scalable Autonomous Driving (CVPR 2024 · 51 citations)
- 2025 · UniPCGC: Towards Practical Point Cloud Geometry Compression via an Efficient Unified Approach (AAAI 2025 · 26 citations)
- 2026 · DeepRAHT: Learning Predictive RAHT for Point Cloud Attribute Compression (AAAI 2026 · 2 citations)
Earliest papers here
- FlowNet3D: Learning Scene Flow in 3D Point Clouds (CVPR 2019 · 526 citations)
- HPLFlowNet: Hierarchical Permutohedral Lattice FlowNet for Scene Flow Estimation on Large-Scale Point Clouds (CVPR 2019 · 236 citations)
- PointFlowNet: Learning Representations for Rigid Motion Estimation From Point Clouds (CVPR 2019 · 131 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
ScanNet · PSNR (higher is better)
6 papers, 2020–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 22.81 | Neural Point Cloud Rendering via Multi-Plane Projection (CVPR 2020 · 82 citations) |
| 2022 | 30.32 | Point-NeRF: Point-Based Neural Radiance Fields (CVPR 2022 · 561 citations) |
KITTI_o · EPE (lower is better)
4 papers, 2021–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 0.105 | Self-Point-Flow: Self-Supervised Scene Flow Estimation From Point Clouds With Optimal Transport and Random Walk (CVPR 2021 · 46 citations) |
| 2022 | 0.102 | RigidFlow: Self-Supervised Scene Flow Learning on Point Clouds by Local Rigidity Prior (CVPR 2022 · 47 citations) |
| 2023 | 0.047 | SCOOP: Self-Supervised Correspondence and Optimization-Based Scene Flow (CVPR 2023 · 33 citations) |
KITTI_t · EPE (lower is better)
4 papers, 2021–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 0.072 | Self-Point-Flow: Self-Supervised Scene Flow Estimation From Point Clouds With Optimal Transport and Random Walk (CVPR 2021 · 46 citations) |
| 2023 | 0.039 | SCOOP: Self-Supervised Correspondence and Optimization-Based Scene Flow (CVPR 2023 · 33 citations) |
ScanNet · LPIPS (lower is better)
4 papers, 2020–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 0.234 | Neural Point Cloud Rendering via Multi-Plane Projection (CVPR 2020 · 82 citations) |
ScanNet · SSIM (higher is better)
4 papers, 2020–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 0.835 | Neural Point Cloud Rendering via Multi-Plane Projection (CVPR 2020 · 82 citations) |
What has been tried
77 papers here report 827 measured changes (a component added, removed or swapped, compared with a baseline): helped 80%, no clear effect 3%, hurt 14%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| use Latent Rendering rather than no forecasting pre-training | nuScenes 3D detection · NDS (%) | 44.11 -> 47.58 | Helped | Visual Point Cloud Forecasting enables Scalable Autonomous Driving (CVPR 2024 · 51 citations) |
| replace the uniform grouping scheme with Uneven Eight grouping | Lossless point cloud compression · CR gain | 0.0% -> -1.8% | Helped | UniPCGC: Towards Practical Point Cloud Geometry Compression via an Efficient Unified Approach (AAAI 2025 · 26 citations) |
| add explicit IFMP to DAE-M | Loot, QP 51 · ΔPSNR (dB) | 0.41 -> 0.52 | Helped | Fast Inter-frame Motion Prediction for Compressed Dynamic Point Cloud Attribute Enhancement (AAAI 2024 · 26 citations) |
| Use AdaDPCC coarse-to-fine inter-frame prediction instead of 3DAWI | Soldier · encoding time | 8.8% of the time needed by 3DAWI | Helped | AdaDPCC: Adaptive Rate Control and Rate-Distortion-Complexity Optimization for Dynamic Point Cloud Compression (AAAI 2025 · 22 citations) |
| Add Intra-frame Spatial Mamba and Inter-frame Temporal Mamba | MSR-Action3D · accuracy | 90.94 -> 92.68 | Helped | Mamba4D: Efficient 4D Point Cloud Video Understanding with Disentangled Spatial-Temporal State Space Models (CVPR 2025 · 19 citations) |
| Replace the lip point-cloud hash grid with a GNN | Head reconstruction · PSNR | 32.704 -> 31.011 | Hurt | PointTalk: Audio-Driven Dynamic Lip Point Cloud for 3D Gaussian-based Talking Head Synthesis (AAAI 2025 · 15 citations) |
| Add LeCE | 8iVFB · BPP | 0.626 -> 0.592; −5.43% | Helped | TopNet: Transformer-Efficient Occupancy Prediction Network for Octree-Structured Point Cloud Geometry Compression (CVPR 2025 · 13 citations) |
| add the feature transfer module | MPEG 8i average · CD | 0.15 -> 0.11 | Helped | patchDPCC: A Patchwise Deep Compression Framework for Dynamic Point Clouds (AAAI 2024 · 10 citations) |
| use point cloud construction and fusion instead of random matrices and vectors | LLFF · quantitative outcomes | Helped | Depth-Guided Robust and Fast Point Cloud Fusion NeRF for Sparse Input Views (AAAI 2024 · 6 citations) | |
| add implicit fields | IGPLAY SCSF · IoU | 47.0 -> 56.5; 9.5% | Helped | Semantic Complete Scene Forecasting from a 4D Dynamic Point Cloud Sequence (AAAI 2024 · 5 citations) |
| replace direct rotation prediction from scratch with normal-guided rotation | OmniObject3D validation split · PSNR | 25.8 -> 30.9 | Helped | GaussianPainter: Painting Point Cloud into 3D Gaussians with Normal Guidance (AAAI 2025 · 3 citations) |
| Add 3D reference points | ShapeNet SRN Chairs · PSNR | 23.30 -> 23.45 | Helped | LeanGaussian: Breaking Pixel or Point Cloud Correspondence in Modeling 3D Gaussians (CVPR 2025 · 3 citations) |
Baseline numbers that disagree
The same baseline on the same benchmark and metric, reported more than 5% apart by different papers here. Training budgets, data splits and reimplementations differ, so state which number you compare with.
| Baseline | Benchmark | Metric | Reported | Papers |
|---|---|---|---|---|
| NeRF | DTU | PSNR | 28.97 – 31.03 | 2 papers, e.g. ICLR 2023 and AAAI 2023 |
| NeRF | NeRF-Synthetic | PSNR | 31 – 33.07 | 2 papers, e.g. NeurIPS 2023 and AAAI 2023 |
What reviewers will expect
Datasets and baselines named in the 17 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- ShapeNet · 3 of 17 papers
- MVUB · 3 of 17 papers
- 8iVFB · 2 of 17 papers
- MSR-Action3D · 2 of 17 papers
- HOI4D · 2 of 17 papers
- ScanNet · 2 of 17 papers
Baselines
- G-PCC · 3 of 17 papers
- PCGCv2 · 2 of 17 papers
- TriplaneGaussian · 2 of 17 papers
- SparsePCGC · 2 of 17 papers
- VoxelDNN · 2 of 17 papers
- MsVoxelDNN · 2 of 17 papers
- OctAttention · 2 of 17 papers
- ECM-OPCC · 2 of 17 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 55 papers; from OpenAlex and the papers' affiliations).
- Peking University · 7 papers
- Shanghai Jiao Tong University · 5 papers
- The Chinese University of Hong Kong · 4 papers
- Simon Fraser University · 4 papers
- Sun Yat-sen University · 3 papers
- Peng Cheng Laboratory · 3 papers
- Tsinghua University · 3 papers
- Anhui University · 2 papers
- Xi’an Jiaotong University · 2 papers
- Carnegie Mellon University · 2 papers
Use it
Download BibTeX of 39 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Point cloud detection and tracking (56 papers)
- LiDAR point cloud compression and completion (60 papers)
Elsewhere in the topic
- Local feature aggregation networks (158 papers)
- Point-language and open-vocabulary 3D (107 papers)
- Equivariant point cloud networks (84 papers)
- Self-supervised point cloud pretraining (73 papers)
- Efficient point sampling architectures (45 papers)
- Adversarial point cloud attacks (37 papers)
- Masked point autoencoders (22 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).