Research map / Detection and segmentation / 3D detection and driving
Rethinking the Spatio-Temporal Alignment of End-to-End 3D Perception
Xiaoyu Li, Peidong Li, Xian Wu, Long Shi, Dedong Liu, Yitao Wu, Jiajia Fu, Dixiao Cui, Lijun Zhao, Lining Sun
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 HAT alignment module | SparseDrive (2nd stage) · nuScenes val | mAP / AMOTA: 41.2 -> 42.5 mAP; 36.9 -> 40.0 AMOTA | Improved |
| add HAT | SparseDrive (2nd stage) · nuScenes val | collision rate (%): 0.123 -> 0.084 | Improved |
| add HAT | DiffusionDrive · nuScenes val | collision rate (%): 0.080 -> 0.042 | Improved |
| add HAT under snow corruption | SparseDrive · nuScenes-C snow | NDS: 34.1 -> 39.1 | Improved |
| add HAT under fog corruption | SparseDrive · nuScenes-C fog | mAP: 37.2 -> 37.1 | No clear effect |
| add HAT to temporal detectors | StreamPETR/Sparse4D/SimPB · nuScenes val | NDS: 57.1 -> 57.8; 56.4 -> 57.3; 58.6 -> 59.0 | Improved |
| add HAT to E2E tracker | ADA-Track · nuScenes val | AMOTA: 38.4 -> 39.7 | Improved |
| add HAT to ADA-Track on test | ADA-Track · nuScenes test | AMOTA: 45.6 -> 46.0 | Improved |
| replace MLN with HAT as STA module (ablation) | StreamPETR · nuScenes val | NDS: 57.1 -> 57.8 | Improved |
| HAT vs LMM STA module (ablation) | StreamPETR · nuScenes val | NDS: 57.5 -> 57.8 | Improved |
| use multiple motion models vs single CV (ablation) | Sparse4D · nuScenes val | NDS: 56.5 -> 57.3 | Improved |
| add turning models CTRV and CTRA (ablation) | Sparse4D · nuScenes val | mAOE: 0.53 -> 0.42 | 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.
- ADA-Track — ADA-Track: End-to-End Multi-Camera 3D Multi-Object Tracking with Alternating Detection and Association (CVPR 2024) · better on 2 of 2 shared results
- SimPB — SimPB: A Single Model for 2D and 3D Object Detection from Multiple Cameras (ECCV 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
- ConsisVLA-4D: Advancing Spatiotemporal Consistency in Efficient 3D-Perception and 4D-Reasoning for Robotic Manipulation (CVPR 2026 · similarity 0.88)
- FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving (NeurIPS 2025 · similarity 0.87)
- DriveSpatial: A Benchmark for Spatiotemporal Intelligence in VLMs for Autonomous Driving (NeurIPS 2026 · similarity 0.87)
- STSBench: A Spatio-temporal Scenario Benchmark for Multi-modal Large Language Models in Autonomous Driving (NeurIPS 2025 · similarity 0.86)
- TraF-Align: Trajectory-aware Feature Alignment for Asynchronous Multi-agent Perception (CVPR 2025 · similarity 0.86)
- RoadSceneBench: A Lightweight Benchmark for Mid-Level Road Scene Understanding (CVPR 2026 · 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.
