Research map / Detection and segmentation / Tracking
Simple Cues Lead to a Strong Multi-Object Tracker
Jenny Seidenschwarz, Guillem Brasó, Víctor Castro Serrano, Ismail Elezi, Laura Leal-Taixé
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 GHOST's reID and linear motion in place of ByteTrack's Kalman Filter | ByteTrack · DanceTrack | HOTA: 47.1 -> 54.0 | Improved |
| Replace Tracktor's baseline reID with GHOST reID | Tracktor · MOT17 | HOTA: 57.7 -> 58.7 | Improved |
| Use mean distance across an inactive track's feature vectors rather than a proxy feature vector (ablation) | GHOST · BDD100k, DanceTrack, MOT17, MOT20 validation | tracking performance stability across datasets | Improved |
| Vary the appearance-motion weight according to dataset motion and occlusion (ablation) | GHOST · BDD100k, DanceTrack, MOT17, MOT20 | HOTA | Mixed |
| Vary the number of frames used for linear velocity estimation by dataset (ablation) | GHOST · BDD100k, DanceTrack, MOT17, MOT20 | HOTA | Mixed |
| Use separate association thresholds for active and inactive tracks (ablation) | GHOST · Multi-object tracking | tracking performance | Improved |
| Keep inactive tracks for 50 frames instead of removing them immediately (ablation) | GHOST · MOT17 validation | HOTA | Improved |
| Apply on-the-fly domain adaptation to distance distributions (ablation) | GHOST · MOT17 validation | stability of association thresholds | Improved |
| Combine appearance and motion cues rather than relying on appearance alone (ablation) | GHOST · MOT17 sequences with low visibility or camera motion | rate of correct associations | 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.
- QDTrack — QDTrack: Quasi-Dense Similarity Learning for Appearance-Only Multiple Object Tracking (TPAMI 2023) · better on 5 of 6 shared results
- ByteTrack — ByteTrack: Multi-Object Tracking by Associating Every Detection Box (ECCV 2022) · better on 4 of 6 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
- DanceTrack: Multi-Object Tracking in Uniform Appearance and Diverse Motion (CVPR 2022 · similarity 0.89)
- Tracking Without Bells and Whistles (ICCV 2019 · similarity 0.89)
- Hybrid-SORT: Weak Cues Matter for Online Multi-Object Tracking (AAAI 2024 · similarity 0.88)
- Multiple Object Tracking as ID Prediction (CVPR 2025 · similarity 0.88)
- DiffusionTrack: Diffusion Model for Multi-Object Tracking (AAAI 2024 · similarity 0.87)
- Video Diffusion Models Excel at Tracking Similar-Looking Objects Without Supervision (NeurIPS 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.
