Research map / 3D vision / 3D generation
SEM-ROVER: Semantic Voxel-Guided Diffusion for Large-Scale Driving Scene Generation
Hiba Dahmani, Nathan Piasco, Moussab Bennehar, Luis G Roldao Jimenez, Dzmitry Tsishkou, Laurent Caraffa, Jean-Philippe Tarel, Roland Brémond
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 conditioning (ablation) | SEM-ROVER diffusion · 3D feature-space ablation | E3D: 3.927 -> 3.523 | Improved |
| add semantic conditioning (ablation) | SEM-ROVER diffusion · 3D feature-space ablation | MMDS: 0.105 -> 0.091 | Improved |
| add point ordering by distance to voxel center (ablation) | SEM-ROVER diffusion · 3D feature-space ablation | E3D: 3.585 -> 3.523 | Improved |
| add point ordering by distance to voxel center (ablation) | SEM-ROVER diffusion · 3D feature-space ablation | MMDS: 0.093 -> 0.091 | Improved |
| use SEM-ROVER instead of GEN3C | SEM-ROVER · WOD+PandaSet seen views | FID: 113.27 -> 81.98 | Improved |
| use SEM-ROVER instead of InfiniCube on novel views | SEM-ROVER · WOD+PandaSet novel views | FID: 99.13 -> 89.20 | Improved |
| use SEM-ROVER instead of InfiniCube on seen views | SEM-ROVER · WOD+PandaSet seen views | FID: 84.14 -> 81.98 | Improved |
| use SEM-ROVER instead of InfiniCube | SEM-ROVER · WOD+PandaSet seen views | KID: 0.03 -> 0.05 | Hurt |
| use SEM-ROVER instead of InfiniCube for memory | SEM-ROVER · Inference | Min. VRAM: 75 GB -> 8 GB | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- GEN3C — GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control (CVPR 2025) · better on 5 of 5 shared results
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 0.91)
- CymbaDiff: Structured Spatial Diffusion for Sketch-based 3D Semantic Urban Scene Generation (NeurIPS 2025 · similarity 0.89)
- UniScene: Unified Occupancy-centric Driving Scene Generation (CVPR 2025 · similarity 0.89)
- SemCity: Semantic Scene Generation with Triplane Diffusion (CVPR 2024 · similarity 0.88)
- A 3D Scene is Worth 1K Tokens: 3D-Grounded Representation for Scene Generation at Scale (NeurIPS 2026 · similarity 0.88)
- Outdoor Scene Extrapolation with Hierarchical Generative Cellular Automata (CVPR 2024 · 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.
