Research map / Label-efficient and robust learning / Recognition and long-tail
ST-P3: End-to-End Vision-Based Autonomous Driving via Spatial-Temporal Feature Learning
Shengchao Hu, Li Chen, Penghao Wu, Hongyang Li, Junchi Yan, Dacheng Tao
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 Egocentric Aligned Accumulation (ablation) | ST-P3 · nuScenes val | vehicle IoU: 38.00 -> 38.79 | Improved |
| add explicit depth supervision (ablation) | ST-P3 · nuScenes val | vehicle IoU: 38.79 -> 40.10 | Improved |
| add Dual Modelling (ablation) | ST-P3 · nuScenes val | vehicle PQ: 28.63 -> 31.35 | Improved |
| add Dual Modelling (ablation) | ST-P3 · nuScenes val | vehicle IoU: 37.09 -> 38.16 | Improved |
| add loss for all timestamps (LFA) (ablation) | ST-P3 · nuScenes val | vehicle PQ: 31.35 -> 31.72 | Improved |
| add loss for all timestamps (LFA) (ablation) | ST-P3 · nuScenes val | vehicle IoU: 38.16 -> 38.63 | Improved |
| add front-view GRU refinement to the sampler (ablation) | ST-P3 · nuScenes val | collision rate (%): 0.850 -> 0.513 | Improved |
| add prior-knowledge sampler to the refinement unit (ablation) | ST-P3 · nuScenes val | collision rate (%): 1.089 -> 0.513 | Improved |
| add front-view GRU refinement to the sampler (ablation) | ST-P3 · nuScenes val | L2 error (m): 2.128 -> 1.890 | Improved |
| use Bernoulli instead of Gaussian future uncertainty | ST-P3 · nuScenes val | future semantic IoU (%): 38.63 -> 38.87 | No clear effect |
| replace ground-truth-only vanilla planner with ST-P3 cost-based planner | ST-P3 · nuScenes open-loop planning | L2 / collision: L2 3s 2.80 -> 2.90; collision 3s 2.76 -> 1.27 | Mixed |
| compare ST-P3 with LiDAR-based Transfuser on long routes | ST-P3 · CARLA Town05 Long | driving score: 33.15 vs 11.45 | Hurt |
Later papers that compared with this method
6 later accepted papers report ST-P3 as a baseline in their results tables; 6 of them report a better number on most of the results they share with it.
- ResWorld: Temporal Residual World Model for End-to-End Autonomous Driving (ICLR 2026 · better on 2 of 2 shared results)
- CorrectAD: A Self-Correcting Agentic System to Improve End-to-end Planning in Autonomous Driving (AAAI 2026 · better on 2 of 2 shared results)
- DiverseAD: A Large-Scale Driving Dataset with Diverse Atmospheric Conditions (ECCV 2026 · better on 1 of 1 shared result)
- SynAD: Enhancing Real-World End-to-End Autonomous Driving Models through Synthetic Data Integration (ICCV 2025 · better on 2 of 2 shared results)
- PPAD: Iterative Interactions of Prediction and Planning for End-to-end Autonomous Driving (ECCV 2024 · better on 1 of 1 shared result)
- VAD: Vectorized Scene Representation for Efficient Autonomous Driving (ICCV 2023 · better on 4 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
- SpaceDrive: Infusing Spatial Awareness into VLM-based Autonomous Driving (CVPR 2026 · similarity 0.87)
- Explicit Geometric Chain-of-Thought for Vision-Language-Action in Autonomous Driving (NeurIPS 2026 · similarity 0.86)
- Future-Aware End-to-End Driving: Bidirectional Modeling of Trajectory Planning and Scene Evolution (NeurIPS 2025 · similarity 0.86)
- SGDrive: Scene-to-Goal Hierarchical World Cognition for Autonomous Driving (CVPR 2026 · similarity 0.86)
- VR-Drive: Viewpoint-Robust End-to-End Driving with Feed-Forward 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.86)
- FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving (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.
