Research map / Point clouds / Point cloud classification
Low-Latency Neural LiDAR Compression with 2D Context Models
Rui Song, Yan Wang, Tongda Xu, Zhening Liu, Zehong Lin, Jun Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add camera context model (ablation) | RangeCM-G · WOD geometry | BD-Rate: +6.85% | Improved |
| add camera context model (ablation) | RangeCM-GI · WOD intensity | BD-Rate: +2.30% | Improved |
| add flow-based temporal context model (ablation) | RangeCM-G · WOD geometry | BD-Rate: +22.02% vs +6.85% | Improved |
| add flow-based temporal context model (ablation) | RangeCM-GI · WOD intensity | BD-Rate: +21.88% vs +2.30% | Improved |
| add multi-scale context model (ablation) | RangeCM-G · WOD geometry | BD-Rate: +34.19% vs +22.02% | Improved |
| add multi-scale context model (ablation) | RangeCM-GI · WOD intensity | BD-Rate: +31.75% vs +21.88% | Improved |
| add refined camera context after decoding sketch map (ablation) | RangeCM-G · WOD geometry | BD-Rate: 5.11% | Improved |
| add camera context with only 2 camera views (ablation) | RangeCM-G · reorganized KITTI geometry | BD-Rate: 4.77% | Improved |
| joint geometry-intensity training (GI instead of G) | RangeCM · WOD geometry | BD-Rate vs G-PCC: -61.96 -> -59.94 | Hurt |
| joint geometry-intensity training (GI instead of G) | RangeCM · SemanticKITTI geometry | BD-Rate vs G-PCC: -56.07 -> -51.56 | Hurt |
| predict intensity from shared hybrid context | RangeCM-GI · SemanticKITTI intensity | coding time vs Unicorn: 14.84 -> 0.10 | Improved |
| fine-tune WOD-pretrained model on KITTI instead of training from scratch | RangeCM-GI · reorganized KITTI | BD-Rate: 3.84% geometry, 3.25% intensity | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- RIDDLE — RIDDLE: Lidar Data Compression With Range Image Deep Delta Encoding (CVPR 2022) · better on 3 of 3 shared results
- RENO — RENO: Real-Time Neural Compression for 3D LiDAR Point Clouds (CVPR 2025) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- RENO: Real-Time Neural Compression for 3D LiDAR Point Clouds (CVPR 2025 · similarity 0.90)
- msLPCC: A Multimodal-Driven Scalable Framework for Deep LiDAR Point Cloud Compression (AAAI 2024 · similarity 0.88)
- Efficient LiDAR Reflectance Compression via Scanning Serialization (ICML 2025 · similarity 0.88)
- Towards Realistic Scene Generation with LiDAR Diffusion Models (CVPR 2024 · similarity 0.87)
- TREND: Unsupervised 3D Representation Learning via Temporal Forecasting for LiDAR Perception (NeurIPS 2025 · similarity 0.87)
- SCP: Spherical-Coordinate-Based Learned Point Cloud Compression (AAAI 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
