Research map / 3D vision / 3D generation
A Self-Conditioned Representation Guided Diffusion Model for Realistic Text-to-LiDAR Scene Generation
Wentao Qu, Guofeng Mei, Yang Wu, YongShun Gong, Xiaoshui Huang, Liang Xiao
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 SCRG and DPE to text-guided T2LDM (ablation) | T2LDM · nuScenes text-guided | FSVD / TBR: FSVD 73.64 -> 66.93; TBR 19.32 -> 23.44 | Improved |
| add SCRG alone (without DPE) (ablation) | T2LDM baseline · nuScenes text-guided | FSVD: 73.64 -> 68.45 | Improved |
| add DPE alone (without SCRG) (ablation) | T2LDM baseline · nuScenes text-guided | FSVD: 73.64 -> 71.32 | Improved |
| add SCRG for faster convergence (ablation) | T2LDM (DPE only) · KITTI-360 at 30k iterations | FSVD: 91.32 -> 47.29 | Improved |
| train guidance network end-to-end rather than pretrained (ablation) | SCRG · nuScenes | FSVD: 68.77 -> 64.21 | Improved |
| pretrain the guidance network instead of no guidance (ablation) | T2LDM (DPE only) · nuScenes | FSVD: 68.11 -> 68.77 | Hurt |
| use explicit location text prompt (ablation) | T2LDM · nuScenes text prompts | TBR (%): TBR 12.23 vs 31.12 for 'Two cars.' | Hurt |
| use longer prompts with similar semantics (ablation) | T2LDM · nuScenes text prompts | FSID / TBR: FSID 67.32 -> 67.63; TBR 31.12 -> 29.88 | Hurt |
| use different prompt form at inference than training (ablation) | T2LDM · nuScenes text prompts | TBR (%): 31.12 -> 29.23 | No clear effect |
| adjust location text to sample-distribution-aware form (ablation) | T2LDM prompts · nuScenes text prompts | TBR (%): 12.23 -> 23.42 | Improved |
| adjust quantity text (less/more than five cars) (ablation) | T2LDM prompts · nuScenes text prompts | TBR (%): 31.12 -> 60.35 | Improved |
| add directional position encoding | T2LDM · nuScenes | street distortion | Improved |
Closest papers by meaning
- Towards Realistic Scene Generation with LiDAR Diffusion Models (CVPR 2024 · similarity 0.91)
- Spiral: Semantic-Aware Progressive LiDAR Scene Generation and Understanding (NeurIPS 2025 · similarity 0.89)
- LiDAR-LLM: Exploring the Potential of Large Language Models for 3D LiDAR Understanding (AAAI 2025 · similarity 0.88)
- VLM2Scene: Self-Supervised Image-Text-LiDAR Learning with Foundation Models for Autonomous Driving Scene Understanding (AAAI 2024 · similarity 0.88)
- TREND: Unsupervised 3D Representation Learning via Temporal Forecasting for LiDAR Perception (NeurIPS 2025 · similarity 0.87)
- LiFi: LiDAR Generation from Multi-View Images via Geometric and Semantic Collaborative Guidance (NeurIPS 2026 · similarity 0.87)
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.
