Research map / Architectures and efficiency / RNNs and normalization
MuSCLE: Multi Sweep Compression of LiDAR using Deep Entropy Models
Sourav Biswas, Jerry Liu, Kelvin Wong, Shenlong Wang, Raquel Urtasun
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 past-sweep top-down aggregated features (T) (ablation) | occupancy model with past occupancy bytes · UrbanCity spatial bitrate (D=12) | spatial bitrate: 2.91 -> 2.87 | Improved |
| add bottom-up aggregated features (B) (ablation) | occupancy model O+T · UrbanCity spatial bitrate (D=12) | spatial bitrate: 2.87 -> 2.72 | Improved |
| add continuous convolution (CC) for spatio-temporal aggregation (ablation) | occupancy model O+T+B · UrbanCity spatial bitrate (D=12) | spatial bitrate: 2.72 -> 2.55 | Improved |
| add continuous convolution (CC) at depth 16 (ablation) | occupancy model O+T+B · UrbanCity spatial bitrate (D=16) | spatial bitrate: 13.95 -> 13.79 | Improved |
| use past-sweep intensity context via MLP instead of zlib (ablation) | intensity model · SemanticKITTI intensity bitrate (D=12) | intensity bitrate: 2.42 -> 2.31 | Improved |
| use continuous convolution over multiple neighbors instead of single-neighbor MLP (ablation) | intensity model · SemanticKITTI intensity bitrate (D=12) | intensity bitrate: 2.31 -> 2.13 | Improved |
| use continuous convolution intensity model vs zlib at depth 16 (ablation) | intensity model · SemanticKITTI intensity bitrate (D=16) | intensity bitrate: 5.23 -> 4.68 | Improved |
| compress leaf offsets with zlib/LZMA/bzip2 (ablation) | octree leaf offsets · UrbanCity leaf offsets | avg bytes per sweep: 102429.31 -> 102468.93 (zlib) | Hurt |
| add temporal context to octree entropy model vs OctSqueeze | MuSCLE · UrbanCity / SemanticKITTI joint bitrate at equal quality | bitrate reduction: 7-17% / 6-19% lower | Improved |
Closest papers by meaning
- OctSqueeze: Octree-Structured Entropy Model for LiDAR Compression (CVPR 2020 · similarity 0.90)
- Efficient LiDAR Reflectance Compression via Scanning Serialization (ICML 2025 · similarity 0.86)
- msLPCC: A Multimodal-Driven Scalable Framework for Deep LiDAR Point Cloud Compression (AAAI 2024 · similarity 0.86)
- RENO: Real-Time Neural Compression for 3D LiDAR Point Clouds (CVPR 2025 · similarity 0.86)
- SCP: Spherical-Coordinate-Based Learned Point Cloud Compression (AAAI 2024 · similarity 0.83)
- TREND: Unsupervised 3D Representation Learning via Temporal Forecasting for LiDAR Perception (NeurIPS 2025 · similarity 0.83)
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.
