Research map / Image generation / Fast diffusion sampling
Adaptive Piecewise Distillation for Efficient LiDAR Data Generation
Ruibo Li, Xiaofeng Yang, Ze Yang, Jiacheng Wei, Chunyan Miao, Guosheng Lin
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 |
|---|---|---|---|
| use adaptive piecewise distillation instead of timestep distillation on 2-RF | R2Flow pixel-based, 4 steps · KITTI-360 unconditional (Range+Reflectance) | FRD: 187.10 -> 165.31 | Improved |
| use adaptive piecewise distillation instead of the reflow operation (ablation) | R2Flow pixel-based, 4 steps · KITTI-360 unconditional (Range+Reflectance) | FRD: 221.87 -> 165.31 | Improved |
| use piecewise reflow instead of reflow at 4 steps (ablation) | R2Flow pixel-based, 4 steps · KITTI-360 unconditional (Range+Reflectance) | FRD: 221.87 -> 357.00 | Hurt |
| add adaptive timestep selection over uniform piecewise intervals (ablation) | piecewise distillation · KITTI-360 unconditional (Range+Reflectance) | FRD: 178.31 -> 165.31 | Improved |
| add adaptive timestep selection over uniform piecewise intervals (ablation) | piecewise distillation · KITTI-360 unconditional (Range only) | FRID: 145.7 -> 144.0 | Improved |
| add adaptive timestep selection over uniform piecewise intervals (ablation) | piecewise distillation · KITTI-360 camera-to-LiDAR | FRID: 27.1 -> 27.0 | Improved |
| add adaptive timestep selection over uniform piecewise intervals (ablation) | piecewise distillation · KITTI-360 camera-to-LiDAR | MMD: 3.40 -> 3.41 | No clear effect |
| use piecewise distillation (uniform) instead of reflow (ablation) | 4-step student · KITTI-360 unconditional (Range only) | FRID: 164.9 -> 145.7 | Improved |
| distill 4-step student from teacher instead of sampling teacher with 4 steps (ablation) | R2Flow 1-RF pixel-based · KITTI-360 unconditional (Range+Reflectance) | FRD: 699.90 -> 165.31 | Improved |
| use 4-step LiDAR-Flow instead of 4-step LiDM | latent-based model · KITTI-360 unconditional (Range only) | FRID: 305.5 -> 144.0 | Improved |
| use 4-step LiDAR-Flow instead of 4-step LiDM | latent-based model · SemanticKITTI semantic-map-to-LiDAR | FRID: 24.8 -> 13.9 | Improved |
| use 4-step LiDAR-Flow instead of 50-step LiDM | latent-based model · SemanticKITTI semantic-map-to-LiDAR | FRID: 12.3 -> 13.9 | Mixed |
Closest papers by meaning
- LiFi: LiDAR Generation from Multi-View Images via Geometric and Semantic Collaborative Guidance (NeurIPS 2026 · similarity 0.87)
- Towards Realistic Scene Generation with LiDAR Diffusion Models (CVPR 2024 · similarity 0.87)
- WeatherGen: A Unified Diverse Weather Generator for LiDAR Point Clouds via Spider Mamba Diffusion (CVPR 2025 · similarity 0.86)
- LiDAR4D: Dynamic Neural Fields for Novel Space-time View LiDAR Synthesis (CVPR 2024 · similarity 0.85)
- Towards foundational LiDAR world models with efficient latent flow matching (NeurIPS 2025 · similarity 0.85)
- D^3CTTA: Domain-Dependent Decorrelation for Continual Test-Time Adaption of 3D LiDAR Segmentation (CVPR 2025 · similarity 0.85)
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.
