Research map / 3D vision / Human pose estimation
Multi-target Tracking in Multiple Non-overlapping Cameras Using Fast-Constrained Dominant Sets
Yonatan Tariku Tesfaye, Eyasu Zemene, Andrea Prati, Marcello Pelillo, Mubarak A. Shah
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 |
|---|---|---|---|
| rank by constrained dominant set membership score instead of pairwise Euclidean distance (ablation) | CDSC on MARS (CNN+XQDA) · MARS across-camera ReID | average precision: 0.42 -> 0.45 | Improved |
| rank by membership score instead of pairwise distance (ablation) | CDSC on MARS (CNN+XQDA) · MARS within-camera ReID | average precision: 0.65 -> 0.68 | Improved |
| rank by membership score instead of pairwise distance (ablation) | CDSC on MARS (CNN+KISSME) · MARS across-camera ReID | average precision: 0.41 -> 0.44 | Improved |
| rank by membership score instead of pairwise distance (ablation) | CDSC on MARS (CNN+Eucl) · MARS across-camera ReID | average precision: 0.29 -> 0.29 | No clear effect |
| use constrained dominant set tracking instead of prior tracker | CDSC tracker · DukeMTMC Test-Easy | MOTA: 59.4 -> 70.9 | Improved |
| use constrained dominant set tracking instead of prior tracker | CDSC tracker · DukeMTMC Test-Hard | MOTA: 54.6 -> 59.6 | Improved |
| use constrained dominant set tracking instead of prior tracker | CDSC tracker · DukeMTMC Test-Easy | identity switches: 300 -> 693 | Hurt |
| use constrained dominant set tracking instead of prior tracker | CDSC tracker · DukeMTMC Test-Hard | identity switches: 652 -> 1637 | Hurt |
| use across-camera constrained dominant sets instead of prior method | CDSC tracker · DukeMTMC Test-Easy multi-camera | IDF1: 56.2 -> 60.0 | Improved |
| use across-camera constrained dominant sets instead of prior method | CDSC tracker · DukeMTMC Test-Hard multi-camera | IDF1: 47.3 -> 50.9 | Improved |
| use constrained dominant sets instead of CNN+XQDA baseline | CDSC ReID · MARS | rank-1 accuracy: 65.30 -> 68.22 | Improved |
| speed up CDSC with fast InfImDyn-based local solver (FCDSC) | CDSC · MARS track association | CPU time ratio: 2000x faster to above 4500x | Improved |
Closest papers by meaning
- Visio-Temporal Attention for Multi-Camera Multi-Target Association (ICCV 2021 · similarity 0.86)
- SHaPE: A Novel Graph Theoretic Algorithm for Making Consensus-Based Decisions in Person Re-Identification Systems (ICCV 2017 · similarity 0.86)
- PoseTrack: Joint Multi-Person Pose Estimation and Tracking (CVPR 2017 · similarity 0.86)
- Multiple People Tracking by Lifted Multicut and Person Re-Identification (CVPR 2017 · similarity 0.86)
- PoseTrack21: A Dataset for Person Search, Multi-Object Tracking and Multi-Person Pose Tracking (CVPR 2022 · similarity 0.86)
- MTMMC: A Large-Scale Real-World Multi-Modal Camera Tracking Benchmark (CVPR 2024 · 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.
