Research map / Detection and segmentation / Tracking
Delving into Motion-Aware Matching for Monocular 3D Object Tracking
Kuan-Chih Huang, Ming-Hsuan Yang, Yi-Hsuan Tsai
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 |
|---|---|---|---|
| use motion representation instead of global representation (feature-space matching) (ablation) | MoMA-M3T · nuScenes val | AMOTA: 28.8 -> 30.7 | Improved |
| match in feature space instead of output space (motion representation) (ablation) | MoMA-M3T · nuScenes val | AMOTA: 28.7 -> 30.7 | Improved |
| add contrastive motion feature learning (ablation) | MoMA-M3T · nuScenes val | AMOTA: 30.7 -> 31.1 | Improved |
| add temporal encoder to motion transformer (ablation) | MoMA-M3T · nuScenes val | AMOTA: 29.5 -> 31.1 | Improved |
| add spatial encoder (ablation) | MoMA-M3T · nuScenes val | AMOTA: 30.1 -> 31.1 | Improved |
| add global positional feature in spatial encoder (ablation) | MoMA-M3T · nuScenes val | AMOTA: 30.7 -> 31.1 | Improved |
| use motion transformer instead of LSTM (ablation) | MoMA-M3T · nuScenes val | AMOTA: 29.7 -> 31.1 | Improved |
| add learnable time positional encoding (ablation) | motion transformer · nuScenes val | AMOTA: 30.7 -> 31.1 | Improved |
| increase number of history frames T from 1 to 6 (ablation) | motion transformer · nuScenes val | AMOTA: 29.5 -> 31.1 | Improved |
| use MoMA-M3T tracker instead of KF3D or LSTM on unseen detectors without retraining | FCOS3D / EPro-PnP detections · nuScenes val | AMOTA: 25.8 -> 29.7 (EPro-PnP KF3D) | Improved |
| MoMA-M3T versus Time3D on AMOTP | nuScenes test single-camera · nuScenes test | AMOTP (m): 1.36 -> 1.479 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Time3D — Time3D: End-to-End Joint Monocular 3D Object Detection and Tracking for Autonomous Driving (CVPR 2022) · better on 1 of 2 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
- M3SOT: Multi-Frame, Multi-Field, Multi-Space 3D Single Object Tracking (AAAI 2024 · similarity 0.89)
- TrackCraft3R: Repurposing Video Diffusion Transformers for Dense 3D Tracking (NeurIPS 2026 · similarity 0.86)
- ADA-Track: End-to-End Multi-Camera 3D Multi-Object Tracking with Alternating Detection and Association (CVPR 2024 · similarity 0.86)
- S2-Track: A Simple yet Strong Approach for End-to-End 3D Multi-Object Tracking (ICML 2025 · similarity 0.86)
- TrackingWorld: World-centric Monocular 3D Tracking of Almost All Pixels (NeurIPS 2025 · similarity 0.86)
- GRAE-3DMOT: Geometry Relation-Aware Encoder for Online 3D Multi-Object Tracking (CVPR 2025 · similarity 0.85)
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.
