Research map / Detection and segmentation / Tracking
Non-Markovian Globally Consistent Multi-Object Tracking
Andrii Maksai, Xinchao Wang, Francois Fleuret, Pascal Fua
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 learned global non-Markovian patterns (supervised) to KSP | KSP · people tracking datasets (average) | IDF1: +0.16 | Improved |
| add learned global patterns (unsupervised) to KSP | KSP · people tracking datasets (average) | IDF1: +0.15 | Improved |
| add learned global patterns (supervised) to MDP | MDP · people tracking datasets (average) | IDF1: +0.05 | Improved |
| add learned global patterns to tracker, effect on MOTA | KSP, MDP, RNN, SORT · people tracking datasets (average) | MOTA: -0.01 to 0.06 | Mixed |
| add supervised patterns to BIPCC on Duke Easy | BIPCC · Duke Easy | IDF1: +1.1% IDF | Improved |
| add supervised patterns to BIPCC on Duke Hard | BIPCC · Duke Hard | IDF1: +0.5% IDF | Improved |
| add unsupervised patterns to BIPCC on Duke trainval-mini | BIPCC · Duke trainval-mini | IDF1: +0.9% IDF | Improved |
| add supervised patterns to DM output on MOT16 | DM · MOT16 (two sequences) | identity switches: 37% drop | Improved |
| add unsupervised patterns to DM output on MOT16 | DM · MOT16 (two sequences) | identity switches: 25% drop | Improved |
| replace learned patterns with a pencil of lines in all directions (ablation) | Ours · tracking datasets | IDF1/MOTA | Hurt |
| replace patterns with a Markovian local smoothness term (ablation) | Ours · tracking datasets | IDF1/MOTA | Hurt |
| use 5 instead of 10 binary search iterations (ablation) | Ours (supervised) · people tracking datasets | IDF1: 5 instead of 10; IDF1 change not more than 1% | No clear effect |
Closest papers by meaning
- Self-Supervised Multi-Object Tracking with Path Consistency (CVPR 2024 · similarity 0.88)
- Tracking Without Bells and Whistles (ICCV 2019 · similarity 0.87)
- TrackFlow: Multi-Object tracking with Normalizing Flows (ICCV 2023 · similarity 0.87)
- Focusing on Tracks for Online Multi-Object Tracking (CVPR 2025 · similarity 0.86)
- Towards Generalizable Multi-Object Tracking (CVPR 2024 · similarity 0.86)
- Samba: Synchronized Set-of-Sequences Modeling for Multiple Object Tracking (ICLR 2025 · similarity 0.86)
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.
