Research map / Detection and segmentation / Tracking
SpOT: Spatiotemporal Modeling for 3D Object Tracking
Colton Stearns, Davis Rempe, Jie Li, Rareș Ambruș, Sergey Zakharov, Vitor Guizilini, Yanchao Yang, Leonidas J. Guibas
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 SSR sequence refinement (boxes and confidences) (ablation) | SpOT · nuScenes val (car) | AMOTA: 84.5 -> 85.1 | Improved |
| add SSR sequence refinement (boxes and confidences) (ablation) | SpOT · nuScenes val (pedestrian) | AMOTA: 81.1 -> 82.5 | Improved |
| add confidence refinement to box refinement (ablation) | SpOT · nuScenes val (pedestrian) | AMOTA: 81.9 -> 82.5 | Improved |
| add box refinement to confidence refinement (ablation) | SpOT · nuScenes val (pedestrian) | AMOTA: 81.6 -> 82.5 | Improved |
| add set abstraction and feature propagation branch to PointNet (ablation) | SpOT · nuScenes val (car) | AMOTA: 84.8 -> 84.8 | No clear effect |
| add self-attention to the encoder (ablation) | SpOT · nuScenes val (car) | AMOTA: 84.8 -> 85.1 | Improved |
| add self-attention to the encoder (ablation) | SpOT · nuScenes val (pedestrian) | AMOTA: 81.7 -> 82.5 | Improved |
| increase history from 10 to 40 sweeps (ablation) | SpOT · nuScenes val (pedestrian) | AMOTA: 81.1 -> 82.5 | Improved |
| increase history from 10 to 40 sweeps (ablation) | SpOT · nuScenes val (car) | AMOTA: 84.7 -> 85.1 | Improved |
| increase history from 1 to 10 sweeps (ablation) | SpOT · Waymo Open val (vehicle) | MOTA: 53.4 -> 55.7 | Improved |
| increase history from 10 to 15 sweeps (ablation) | SpOT · Waymo Open val (vehicle) | MOTA: 55.7 -> 54.5 | Hurt |
| add SSR module on Waymo pedestrians | SpOT · Waymo Open val (pedestrian) | MOTA: 56.5 -> 60.5 | Improved |
Closest papers by meaning
- Standing Between Past and Future: Spatio-Temporal Modeling for Multi-Camera 3D Multi-Object Tracking (CVPR 2023 · similarity 0.90)
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · similarity 0.89)
- M3SOT: Multi-Frame, Multi-Field, Multi-Space 3D Single Object Tracking (AAAI 2024 · similarity 0.88)
- TrackIME: Enhanced Video Point Tracking via Instance Motion Estimation (NeurIPS 2024 · similarity 0.87)
- Tracktention: Leveraging Point Tracking to Attend Videos Faster and Better (CVPR 2025 · similarity 0.87)
- TrackCraft3R: Repurposing Video Diffusion Transformers for Dense 3D Tracking (NeurIPS 2026 · 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.
