Research map / Architectures and efficiency / Pruning and efficient tuning
ELiC: Efficient LiDAR Geometry Compression via Cross-Bit-depth Feature Propagation and Bag-of-Encoders
Junsik Kim, Gun Bang, Soowoong Kim
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 cross-bit-depth feature propagation (ablation) | RENO-style coding network · Ford D1 BD-rate vs RENO | BD-rate (%): -8.60 | Improved |
| add cross-bit-depth feature propagation (ablation) | RENO-style coding network · SemanticKITTI D1 BD-rate vs RENO | BD-rate (%): -6.61 | Improved |
| add cross-bit-depth feature propagation (ablation) | RENO-style coding network · Runtime | encoding time ratio: 0.90x enc, 0.85x dec vs RENO | Improved |
| add Bag-of-Encoders with K=3 (ablation) | ELiC w/o BoE · Ford D1 BD-rate | BD-rate (%): -1.31 | Improved |
| add Bag-of-Encoders with K=3 (ablation) | ELiC w/o BoE · SemanticKITTI D1 BD-rate | BD-rate (%): -2.89 | Improved |
| increase BoE pool size from K=5 to K=7 (ablation) | ELiC · Ford D1 BD-rate | BD-rate (%): -1.98 -> -1.43 | Hurt |
| increase BoE pool size from K=5 to K=6 (ablation) | ELiC · SemanticKITTI D1 BD-rate | BD-rate (%): -3.54 -> -3.53 | No clear effect |
| use 14 dedicated per-level networks instead of K=5 pool (ablation) | ELiC · Ford D1 BD-rate | BD-rate (%): -1.98 -> -2.52 | Improved |
| use Morton-order-preserving hierarchy instead of explicit per-level sorting (ablation) | ELiC · Runtime | average encoding time (s): 0.142 -> 0.121 | Improved |
| use Morton-order-preserving hierarchy instead of explicit per-level sorting (ablation) | ELiC · Runtime | average decoding time (s): 0.128 -> 0.111 | Improved |
| add asynchronous arithmetic encoding | ELiC · Runtime | average encoding time (s): 0.121 -> 0.106 | Improved |
| double feature channels (ELiC to ELiC-Large) | ELiC · Ford D1 BD-rate | BD-rate (%): -22.97 -> -26.54 | Improved |
Closest papers by meaning
- BitRIC: Efficient Neural Compression of LiDAR Range Images via Hierarchical Bitplanes (ECCV 2026 · similarity 0.91)
- RENO: Real-Time Neural Compression for 3D LiDAR Point Clouds (CVPR 2025 · similarity 0.87)
- msLPCC: A Multimodal-Driven Scalable Framework for Deep LiDAR Point Cloud Compression (AAAI 2024 · similarity 0.86)
- Towards Realistic Scene Generation with LiDAR Diffusion Models (CVPR 2024 · similarity 0.86)
- Flexible Depth Completion for Sparse and Varying Point Densities (CVPR 2024 · similarity 0.86)
- Efficient LiDAR Reflectance Compression via Scanning Serialization (ICML 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.
