Research map / Detection and segmentation / Tracking
Standing Between Past and Future: Spatio-Temporal Modeling for Multi-Camera 3D Multi-Object Tracking
Ziqi Pang, Jie Li, Pavel Tokmakov, Dian Chen, Sergey Zagoruyko, Yu-Xiong Wang
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 |
|---|---|---|---|
| Combine past and future reasoning (ablation) | PF-Track · nuScenes validation | AMOTA: 0.368 -> 0.408 | Improved |
| Add past query refinement using historical queries (ablation) | PF-Track baseline · nuScenes validation | AMOTA: 0.368 -> 0.378 | Improved |
| Add past track refinement to query refinement (ablation) | PF-Track with query refinement · nuScenes validation | IDS: 453 -> 400 | Improved |
| Add future motion prediction and propagate query positions (ablation) | PF-Track baseline · nuScenes validation | AMOTA: 0.368 -> 0.374 | Improved |
| Use predicted long-term trajectories to replace low-confidence localizations (ablation) | PF-Track with motion prediction · nuScenes validation | IDS: 469 -> 155 | Improved |
| Enable track extension (ablation) | PF-Track with 2.0s prediction · nuScenes validation | AMOTA: 0.392 -> 0.402 | Improved |
| Enable track extension (ablation) | PF-Track with 3.0s prediction · nuScenes validation | AMOTA: 0.392 -> 0.402 | Improved |
| Enable track extension (ablation) | PF-Track with 4.0s prediction · nuScenes validation | AMOTA: 0.391 -> 0.408 | Improved |
| Increase prediction horizon from 2.0s to 4.0s with extension enabled (ablation) | PF-Track · nuScenes validation | AMOTA: 0.402 -> 0.408 | Improved |
| Predict trajectories end-to-end from object features rather than track states | PF-Track · nuScenes validation, 4.0s | ADE: 2.01 -> 1.88 | Improved |
| Use end-to-end tracking rather than tuned tracking-by-detection | PF-Track · nuScenes validation | IDS: 2053 -> 166 | Improved |
Closest papers by meaning
- S2-Track: A Simple yet Strong Approach for End-to-End 3D Multi-Object Tracking (ICML 2025 · similarity 0.87)
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · similarity 0.87)
- Focusing on Tracks for Online Multi-Object Tracking (CVPR 2025 · similarity 0.87)
- M3SOT: Multi-Frame, Multi-Field, Multi-Space 3D Single Object Tracking (AAAI 2024 · similarity 0.87)
- DreamTrack: Dreaming the Future for Multimodal Visual Object Tracking (CVPR 2025 · similarity 0.87)
- Tracktention: Leveraging Point Tracking to Attend Videos Faster and Better (CVPR 2025 · similarity 0.87)
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.
