Research map / Point clouds / Point cloud classification
LiFlow: Flow Matching for 3D LiDAR Scene Completion
Andrea Matteazzi, Dietmar Tutsch
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 flow matching (LiFlow) instead of DDPM-based LiDiff | point-level LiDAR scene completion · SemanticKITTI val (seq 08) | Chamfer Distance (m): 0.434 -> 0.309 (no refinement); 0.375 -> 0.228 (with refinement) | Improved |
| LiFlow vs LiDPM on fine-grained Voxel IoU 0.1m with refinement | LiFlow · SemanticKITTI val (seq 08) | Voxel IoU 0.1m (%): 14.9 (LiDPM) vs 12.9 (LiFlow) | Hurt |
| add Chamfer Distance Matching loss (lambda_CDM 0 -> 0.1) to NFM (ablation) | LiFlow · Apollo seq 00 | Chamfer Distance (m): 0.411 -> 0.359 | Improved |
| add CDM loss (lambda_CDM 0 -> 0.1) to NFM (ablation) | LiFlow · Apollo seq 00 | Voxel IoU 0.5m (%): 29.0 -> 27.2 | Hurt |
| add NFM loss to CDM-only training (ablation) | LiFlow · Apollo seq 00 | Voxel IoU 0.5m (%): 21.6 -> 27.2 | Improved |
| raise lambda_CDM from 0.1 to 1 (equal weighting) (ablation) | LiFlow · Apollo seq 00 | Chamfer Distance (m): 0.359 -> 0.355 | No clear effect |
| replace BatchNorm with InstanceNorm in MinkUNet (ablation) | LiFlow · Apollo seq 00 | Chamfer Distance (m): 0.405 -> 0.359 | Improved |
| replace BatchNorm with InstanceNorm in MinkUNet (with refinement) (ablation) | LiFlow · Apollo seq 00 | Voxel IoU 0.1m (%): 7.0 -> 11.1 | Improved |
| add refinement network to LiFlow | LiFlow · SemanticKITTI val (seq 08) | Voxel IoU 0.1m (%): 3.8 -> 12.9 | Improved |
| increase generation steps from 5 to 50 (ablation) | LiFlow · Apollo seq 00 | Chamfer Distance (m): 0.351 -> 0.367 | No clear effect |
| LiFlow inference time vs LiDiff at 50 steps | LiFlow · Apollo seq 00 | Inference time (s): 27.1 -> 28.8 | No clear effect |
Closest papers by meaning
- Scaling Diffusion Models to Real-World 3D LiDAR Scene Completion (CVPR 2024 · similarity 0.90)
- ICP-Flow: LiDAR Scene Flow Estimation with ICP (CVPR 2024 · similarity 0.88)
- VoteFlow: Enforcing Local Rigidity in Self-Supervised Scene Flow (CVPR 2025 · similarity 0.87)
- 3DSFLabelling: Boosting 3D Scene Flow Estimation by Pseudo Auto-labelling (CVPR 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)
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.
