Research map / Point clouds / Point cloud classification
PACE: Post-Causal Entropy Modeling for Learned LiDAR Point Cloud Compression
Jiahao Zhu, Kang You, Dandan Ding, Zhan Ma
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 graph-based positional encoding (ablation) | PACE-FP · Ford | BD-BR (%): -38.22% -> -50.19% | Improved |
| add graph-based positional encoding (ablation) | PACE-AR · Ford | BD-BR (%): -46.14% -> -55.07% | Improved |
| use post-causal instead of fully-causal modeling (ablation) | PACE-AR · Ford | BD-BR (%): -52.34% -> -55.07% | Improved |
| use post-causal instead of fully-causal modeling (ablation) | PACE-AR · Ford | decoding latency at L=14: ~90x slower for fully-causal | Improved |
| increase decoding stages from 1 to 16 (ablation) | PACE · Ford | BD-BR (%): -50.19% -> -54.23% | Improved |
| increase decoding stages from 1 to 16 (ablation) | PACE · Ford | decoding time (s): 1.55 s -> 2.18 s | Hurt |
| use Mamba predictor instead of MLP (ablation) | PACE-AR · Ford | BD-BR (%): -52.49% -> -55.07% | Improved |
| use Mamba predictor instead of Attention (ablation) | PACE-AR · Ford | BD-BR / memory / decoding: -54.83% -> -55.07%, 2.68GB -> 0.58GB, 40.95s -> 16.36s | Mixed |
| use joint predictor instead of separate predictors (ablation) | PACE-FP · Ford | BD-BR (%): -50.68% -> -50.19% | No clear effect |
| use joint predictor instead of separate predictors (ablation) | PACE-AR · Ford | BD-BR (%): -55.93% -> -55.07% | No clear effect |
| use intrinsic-aware preprocessing instead of spherical or Cartesian octree (ablation) | PACE · Ford | BD-BR (%) | Improved |
| reduce point density by downsampling 4x | PACE-FP vs L3C2 · Ford | BD-BR (%): -50.19% -> -30.00% (L3C2 -45.67% -> -29.79%) | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- OctAttention — OctAttention: Octree-Based Large-Scale Contexts Model for Point Cloud Compression (AAAI 2022) · better on 1 of 1 shared result
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
- Towards Practical Lossless Neural Compression for LiDAR Point Clouds (ECCV 2026 · similarity 0.90)
- RENO: Real-Time Neural Compression for 3D LiDAR Point Clouds (CVPR 2025 · similarity 0.88)
- msLPCC: A Multimodal-Driven Scalable Framework for Deep LiDAR Point Cloud Compression (AAAI 2024 · similarity 0.87)
- SCP: Spherical-Coordinate-Based Learned Point Cloud Compression (AAAI 2024 · similarity 0.86)
- Efficient LiDAR Reflectance Compression via Scanning Serialization (ICML 2025 · similarity 0.86)
- TREND: Unsupervised 3D Representation Learning via Temporal Forecasting for LiDAR Perception (NeurIPS 2025 · similarity 0.86)
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.
