Research map / 3D vision / 3D generation
DrivePTS: A Progressive Learning Framework with Textual and Structural Enhancement for Driving Scene Generation
Zhechao Wang, Yiming Zeng, Lufan Ma, Zeqing Fu, Chen Bai, Dongshuo Yin, Ziyao Lin, Cheng Lu
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 multi-view hierarchical VLM descriptions to the vanilla model (ablation) | DrivePTS (T2I-Adapter, SD 2.1) · nuScenes val | FID: 15.10 -> 12.03 | Improved |
| add multi-view hierarchical descriptions (ablation) | DrivePTS · nuScenes val | Road mIoU: 59.77 -> 61.22 | Improved |
| add frequency-guided structure loss to the vanilla model (ablation) | DrivePTS · nuScenes val | Road mIoU: 59.77 -> 62.92 | Improved |
| add frequency-guided structure loss to the vanilla model (ablation) | DrivePTS · nuScenes val | FID: 15.10 -> 14.47 | Improved |
| combine descriptions and frequency loss (ablation) | DrivePTS · nuScenes val | FID: 12.03 -> 11.68 | Improved |
| add mutual information constraint on top of MHD and FGSL (ablation) | DrivePTS · nuScenes val | Vehicle mIoU: 27.16 -> 27.82 | Improved |
| add mutual information constraint on top of MHD and FGSL (ablation) | DrivePTS · nuScenes val | FID: 11.68 -> 11.45 | Improved |
| raise lambda_freq from 0 to 0.5 (Stage 1) (ablation) | DrivePTS · nuScenes val | Road mIoU: 61.22 -> 63.60 | Improved |
| raise lambda_freq to 1 (too large) (ablation) | DrivePTS · nuScenes val | Road mIoU: 63.60 -> 61.95 | Hurt |
| raise lambda_MI from 0.05 to 0.2 (ablation) | DrivePTS · nuScenes val | Road mIoU: 63.95 -> 62.65 | Hurt |
| vary alternating iterative step interval in Stage 1 (250 to 3200) (ablation) | DrivePTS · nuScenes val | Road/vehicle controllability | Mixed |
| use DrivePTS synthetic val data to augment BEV segmentation training | CVT · nuScenes test | Road mIoU: 65.83 -> 67.49 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PerLDiff — PerLDiff: Controllable Street View Synthesis Using Perspective-Layout Diffusion Model (ICCV 2025) · better on 5 of 5 shared results
- MagicDrive — MagicDrive: Street View Generation with Diverse 3D Geometry Control (ICLR 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- X-Scene: Large-Scale Driving Scene Generation with High Fidelity and Flexible Controllability (NeurIPS 2025 · similarity 0.90)
- Generating Multimodal Driving Scenes via Next-Scene Prediction (CVPR 2025 · similarity 0.89)
- SceneCrafter: Controllable Multi-View Driving Scene Editing (CVPR 2025 · similarity 0.89)
- Genesis: Multimodal Driving Scene Generation with Spatio-Temporal and Cross-Modal Consistency (NeurIPS 2025 · similarity 0.89)
- DriveScape: High-Resolution Driving Video Generation by Multi-View Feature Fusion (CVPR 2025 · similarity 0.88)
- UniScene: Unified Occupancy-centric Driving Scene Generation (CVPR 2025 · similarity 0.88)
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.
