Research map / Point clouds / Point cloud classification
LiDAR point cloud compression and completion
Compress, upsample and complete LiDAR point clouds with learned models.
60 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Point cloud classification, keyword cluster “motion · estimation · scene”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
19 papers in 2023–24 and 31 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 8.2% to 12.0%.
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
- 2016 · Removing Clouds and Recovering Ground Observations in Satellite Image Sequences via Temporally Contiguous Robust Matrix Completion (CVPR 2016 · 35 citations)
- 2019 · Dense Depth Posterior (DDP) From Single Image and Sparse Range (CVPR 2019 · 137 citations)
- 2020 · LiDARsim: Realistic LiDAR Simulation by Leveraging the Real World (CVPR 2020 · 239 citations)
- 2021 · PWCLO-Net: Deep LiDAR Odometry in 3D Point Clouds Using Hierarchical Embedding Mask Optimization (CVPR 2021 · 85 citations)
- 2022 · Beyond 3D Siamese Tracking: A Motion-Centric Paradigm for 3D Single Object Tracking in Point Clouds (CVPR 2022 · 108 citations)
- 2023 · LidarGait: Benchmarking 3D Gait Recognition With Point Clouds (CVPR 2023 · 116 citations)
- 2024 · TransLoc4D: Transformer-based 4D Radar Place Recognition (CVPR 2024 · 27 citations)
- 2025 · RENO: Real-Time Neural Compression for 3D LiDAR Point Clouds (CVPR 2025 · 22 citations)
- 2026 · CompTrack: Information Bottleneck-Guided Low-Rank Dynamic Token Compression for Point Cloud Tracking (AAAI 2026 · 2 citations)
Earliest papers here
- Removing Clouds and Recovering Ground Observations in Satellite Image Sequences via Temporally Contiguous Robust Matrix Completion (CVPR 2016 · 35 citations)
- Dense Depth Posterior (DDP) From Single Image and Sparse Range (CVPR 2019 · 137 citations)
- LiDARsim: Realistic LiDAR Simulation by Leveraging the Real World (CVPR 2020 · 239 citations)
What has been tried
42 papers here report 428 measured changes (a component added, removed or swapped, compared with a baseline): helped 81%, no clear effect 3%, hurt 14%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add point cloud refinement | NYL-Night, NYL-Rain, SRC-Night, and SJTU4DPR-TestA · Recall@1 | 93.2 -> 96.6; 75.8 -> 81.0; 86.2 -> 89.0; 84.6 -> 88.6 | Helped | TransLoc4D: Transformer-based 4D Radar Place Recognition (CVPR 2024 · 27 citations) |
| Use 1 × 4 row patches instead of 4 × 4 patches | KITTI 4x · MAE | 0.7138 -> 0.4814 | Helped | TULIP: Transformer for Upsampling of LiDAR Point Clouds (CVPR 2024 · 24 citations) |
| Omit scale-wise feature inheritance | KITTI 14 bit · D1 BD-BR (%) | -12.26 -> -6.06 | Hurt | RENO: Real-Time Neural Compression for 3D LiDAR Point Clouds (CVPR 2025 · 22 citations) |
| Use spherical coordinates instead of Cartesian coordinates | SemanticKITTI and Ford · D1 PSNR | Helped | SCP: Spherical-Coordinate-Based Learned Point Cloud Compression (AAAI 2024 · 18 citations) | |
| replace uniform LiDAR scan simulation with conditional latent diffusion scanning | Synthetic LiDAR scanning · FID | 140.3 -> 53.9 | Helped | BWFormer: Building Wireframe Reconstruction from Airborne LiDAR Point Cloud with Transformer (CVPR 2025 · 10 citations) |
| share geometry pools across sensors instead of sharing them within each sensor | S3DIS Area 5 · mIoU | 69.5 -> 61.4(−8.1) | Hurt | GeoAuxNet: Towards Universal 3D Representation Learning for Multi-sensor Point Clouds (CVPR 2024 · 10 citations) |
| Add the multimodal fusion module to aligned depth and segmentation features | SemanticKITTI · PSNR | Helped | msLPCC: A Multimodal-Driven Scalable Framework for Deep LiDAR Point Cloud Compression (AAAI 2024 · 9 citations) | |
| Add global-local augmentation to FLOT | LiDAR KITTI · EPE3D | 0.0601 -> 0.0189 | Helped | 3DSFLabelling: Boosting 3D Scene Flow Estimation by Pseudo Auto-labelling (CVPR 2024 · 9 citations) |
| add hybrid attention | KITTI mean · Success | 68.9 -> 70.8 | Helped | Modeling Continuous Motion for 3D Point Cloud Object Tracking (AAAI 2024 · 8 citations) |
| use multiple BEV density images rather than a single BEV image | Wild-Places V-03 · Recall@1 | 53.70 -> 76.53 | Helped | ForestLPR: LiDAR Place Recognition in Forests Attentioning Multiple BEV Density Images (CVPR 2025 · 7 citations) |
| add learnable MDP | Seeing Through Fog snow heavy test split · FPD | 84.58 -> 59.28 | Helped | WeatherGen: A Unified Diverse Weather Generator for LiDAR Point Clouds via Spider Mamba Diffusion (CVPR 2025 · 7 citations) |
| Add the Offset Head | SemanticKITTI 08, 09, and 10 · APE | 3.69 -> 4.7; 2.84 -> 2.53; 63.26 -> 1.35 | Mixed | DeepPointMap: Advancing LiDAR SLAM with Unified Neural Descriptors (AAAI 2024 · 7 citations) |
What reviewers will expect
Datasets and baselines named in the 18 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- nuScenes · 5 of 18 papers
- KITTI · 3 of 18 papers
- KITTI-360 · 3 of 18 papers
- Wild-Places · 2 of 18 papers
- Ford · 2 of 18 papers
- Waymo · 2 of 18 papers
Baselines
- TransLoc3D · 2 of 18 papers
- MinkLoc3Dv2 · 2 of 18 papers
- LoGG3D-Net · 2 of 18 papers
- Unicorn · 2 of 18 papers
- LiDARGen · 2 of 18 papers
- LiDARVAE · 2 of 18 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 50 papers; from OpenAlex and the papers' affiliations).
- Shanghai Jiao Tong University · 3 papers
- Tongji University · 3 papers
- Xiamen University · 3 papers
- Nanjing University · 3 papers
- Nanyang Technological University · 2 papers
- Nanjing University of Science and Technology · 2 papers
- Wuhan University · 2 papers
- Waabi · 2 papers
- University of Toronto · 2 papers
- Queensland University of Technology · 2 papers
Use it
Download BibTeX of 33 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
- Scene flow and dynamic point clouds (89 papers)
- Point cloud detection and tracking (56 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).