Research map / 3D vision / 3D generation
Neural Volumetric World Models for Autonomous Driving
Zanming Huang, Jimuyang Zhang, Eshed Ohn-Bar
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 |
|---|---|---|---|
| replace BEV encoder with volumetric representation, both trained from scratch (ablation) | planner · nuScenes val | average collision rate (%): 0.60 -> 0.59 | Improved |
| add self-supervised volumetric pre-training (ablation) | NeMo · nuScenes val | average collision rate (%): 0.59 -> 0.54 | Improved |
| add occupancy prediction to pre-training (ablation) | NeMo pre-training · nuScenes val | average collision rate (%): 0.61 -> 0.54 | Improved |
| add motion flow module to pre-training (ablation) | NeMo pre-training · nuScenes val | average collision rate (%): 0.58 -> 0.54 | Improved |
| add image reconstruction to pre-training (ablation) | NeMo pre-training · nuScenes val | average L2 error (m): 0.98 -> 1.00 | Mixed |
| add temporal attention fusion of predicted future features (ablation) | NeMo planner · nuScenes val | average L2 error (m): 1.00 -> 0.91 | Improved |
| use temporal attention instead of simple concatenation (ablation) | NeMo planner · nuScenes val | average L2 error (m): 0.94 -> 0.91 | Improved |
| add auxiliary reconstruction and occupancy losses during planner fine-tuning (ablation) | NeMo planner · nuScenes val | average L2 error (m): 0.91 -> 0.84 | Improved |
| add self-supervised pre-training (ablation) | NeMo · CARLA Town05 Long | Driving Score: 34.23 -> 42.57 | Improved |
Closest papers by meaning
- Semi-Supervised Vision-Centric 3D Occupancy World Model for Autonomous Driving (ICLR 2025 · similarity 0.87)
- VR-Drive: Viewpoint-Robust End-to-End Driving with Feed-Forward 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.87)
- Driving in the Occupancy World: Vision-Centric 4D Occupancy Forecasting and Planning via World Models for Autonomous Driving (AAAI 2025 · similarity 0.87)
- Achieving Speed-Accuracy Balance in Vision-based 3D Occupancy Prediction via Geometric-Semantic Disentanglement (AAAI 2025 · similarity 0.86)
- GaussianWorld: Gaussian World Model for Streaming 3D Occupancy Prediction (CVPR 2025 · similarity 0.86)
- Future-Aware End-to-End Driving: Bidirectional Modeling of Trajectory Planning and Scene Evolution (NeurIPS 2025 · 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.
