Research map / Detection and segmentation / Tracking
LA-MOTR: End-to-End Multi-Object Tracking by Learnable Association
Peng Wang, Yongcai Wang, Hualong Cao, Wang Chen, Deying Li
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 |
|---|---|---|---|
| replace single shared decoder with decoupled SGLA association (ablation) | LA-MOTR · DanceTrack val (association ablation) | HOTA: 60.220 -> 64.474 | Improved |
| add relative spatial cues to SGLA (ablation) | LA-MOTR · DanceTrack val (association ablation) | HOTA: 62.753 -> 64.474 | Improved |
| replace hand-crafted combination association with learnable SGLA (ablation) | LA-MOTR · DanceTrack val (association ablation) | HOTA: 57.920 -> 64.474 | Improved |
| use concatenated self-attention instead of cross-attention with Eobj as query (ablation) | SGLA · DanceTrack val (SGLA attention ablation) | HOTA: 63.896 -> 64.474 | Improved |
| use concatenated self-attention instead of cross-attention with Etck as query (ablation) | SGLA · DanceTrack val (SGLA attention ablation) | HOTA: 64.008 -> 64.474 | Improved |
| use parameter-free object/tracklet blend instead of learned TAN update (ablation) | LA-MOTR · DanceTrack val (tracklet query update ablation) | HOTA: 62.740 -> 64.474 | Improved |
| use parameter-free blend instead of TIM long-term memory update (ablation) | LA-MOTR · DanceTrack val (tracklet query update ablation) | HOTA: 64.589 -> 64.474 | No clear effect |
| add CrowdHuman data to MOT17 training | LA-MOTR · MOT17 test | HOTA: 57.4 -> 62.6 | Improved |
| use DAB-Deformable-DETR instead of standard Deformable-DETR | LA-MOTR · DanceTrack test | HOTA: 66.5 -> 71.1 | Improved |
| use DAB-Deformable-DETR instead of standard Deformable-DETR | LA-MOTR · SportsMOT test | HOTA: 69.5 -> 72.4 | Improved |
| tune tracklet query update weight w per dataset (ablation) | LA-MOTR · DanceTrack, SportsMOT, MOT17 | HOTA: w = 1, 0.9, 0.5 | Mixed |
| add learnable association module | LA-MOTR vs MOTR · Efficiency | FPS: 13.47 -> 11.96 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- MOTRv3 — MOTR: End-to-End Multiple-Object Tracking with TRansformer (ECCV 2022) · better on 1 of 2 shared results
- MeMOTR — MeMOTR: Long-Term Memory-Augmented Transformer for Multi-Object Tracking (ICCV 2023) · 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
- ADA-Track: End-to-End Multi-Camera 3D Multi-Object Tracking with Alternating Detection and Association (CVPR 2024 · similarity 0.89)
- Multiple Object Tracking as ID Prediction (CVPR 2025 · similarity 0.89)
- Focusing on Tracks for Online Multi-Object Tracking (CVPR 2025 · similarity 0.89)
- DeconfuseTrack: Dealing with Confusion for Multi-Object Tracking (CVPR 2024 · similarity 0.88)
- More Than Meets the Eye: Enhancing Multi-Object Tracking Even with Prolonged Occlusions (ICML 2025 · similarity 0.87)
- OVTR: End-to-End Open-Vocabulary Multiple Object Tracking with Transformer (ICLR 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.
