Research map / Point clouds / Segmentation and detection
SG-LDM: Semantic-Guided LiDAR Generation via Latent-Aligned Diffusion
Zhengkang Xiang, Zizhao Li, Amir Khodabandeh, Kourosh Khoshelham
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 semantic alignment to classifier-free guidance (ablation) | SG-LDM · SemanticKITTI val | FRID: 10.4 -> 4.4 | Improved |
| add classifier-free guidance without semantic alignment (ablation) | SG-LDM · SemanticKITTI val | FSVD: 10.8 -> 32.1 | Hurt |
| add CFG with semantic alignment over no CFG (ablation) | SG-LDM · SemanticKITTI val | FRID: 9.4 -> 4.4 | Improved |
| raise unconditional probability P_uncon from 0.2 to 0.5 (ablation) | SG-LDM · SemanticKITTI val | FRID: 4.4 -> 14.5 | Hurt |
| vary P_uncon between 0.1 and 0.2 (ablation) | SG-LDM · SemanticKITTI val | FRID: 4.5 -> 4.4 | No clear effect |
| use 50% noising in lidar translation instead of 0% or 100% (ablation) | SG-LDM lidar translation · SemanticKITTI segmentation (MinkUNet) | mIoU | Improved |
| use pixel-space SG-LDM instead of latent diffusion LiDM | SG-LDM · SemanticKITTI val | inference speed: 85.53 -> 31.68 steps/s | Hurt |
| add LiDM-generated data to segmentation training | MinkUNet · SemanticKITTI segmentation | mIoU | Hurt |
| add SG-LDM-generated data instead of SynLiDAR | MinkUNet · SemanticKITTI segmentation | mIoU: 0.4% lower than SynLiDAR | No clear effect |
Closest papers by meaning
- A Self-Conditioned Representation Guided Diffusion Model for Realistic Text-to-LiDAR Scene Generation (CVPR 2026 · similarity 0.89)
- Spiral: Semantic-Aware Progressive LiDAR Scene Generation and Understanding (NeurIPS 2025 · similarity 0.88)
- Towards foundational LiDAR world models with efficient latent flow matching (NeurIPS 2025 · similarity 0.88)
- 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)
- Construct to Associate: Cooperative Context Learning for Domain Adaptive Point Cloud Segmentation (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.
