Research map / Detection and segmentation / Tracking
Deep Network Flow for Multi-Object Tracking
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Learn nonlinear two-layer MLP pairwise costs end-to-end instead of using hand-crafted costs | Network-flow tracking · KITTI-Tracking cross-validation | MOTA: 73.75 -> 74.19 | Helped |
| Learn nonlinear two-layer MLP pairwise costs end-to-end instead of using hand-crafted costs | Network-flow tracking · MOT16 cross-validation | MOTA: 29.19 -> 31.10 | Helped |
| Use a two-layer MLP rather than a linear pairwise-cost model (ablation) | Learned network-flow tracker · MOT16 cross-validation | MOTA: 28.25 -> 31.10 | Helped |
| Add ALFD motion features to learned bounding-box and overlap inputs (ablation) | Learned network-flow tracker · KITTI-Tracking cross-validation | MOTA: 74.12 -> 74.11 | No clear effect |
| Add ALFD motion features to learned bounding-box and overlap inputs (ablation) | Learned network-flow tracker · KITTI-Tracking cross-validation | IDS: 55 -> 29 | Helped |
| Add RGB appearance features to unary costs (ablation) | Learned network-flow tracker with ACF detections · MOT15 cross-validation | MOTA: 28.94 -> 39.08 | Helped |
| Add RGB appearance features to pairwise costs as well as unary costs (ablation) | Learned network-flow tracker with ACF detections · MOT15 cross-validation | IDS: 285 -> 233 | Helped |
| Reduce the weight on ambiguous cases (ablation) | Learned network-flow tracker · KITTI-Tracking cross-validation | REC: 82.84 -> 84.11 | Helped |
| Set the precision-recall loss weight to 0.3 rather than using unweighted loss (ablation) | Learned network-flow tracker · KITTI-Tracking cross-validation | MOTA: 74.07 -> 66.84 | Hurt |
| Increase the link-variable loss weight to 2.0 rather than using unweighted loss (ablation) | Learned network-flow tracker · KITTI-Tracking cross-validation | IDS: 53 -> 42 | Helped |
| Use RGB appearance in unary costs on the benchmark test set | Network-flow tracker with ACF detections · MOT15 benchmark | MOTA: 25.2 -> 26.8 | Helped |
| Use the method with RGB pairwise costs rather than unary-only RGB | MOT15 benchmark tracker · MOT15 benchmark | MOTA: 26.8 -> 29.0 | Hurt |
Closest papers by meaning
- Learning a Neural Solver for Multiple Object Tracking (CVPR 2020 · similarity 0.90)
- Learning of Global Objective for Network Flow in Multi-Object Tracking (CVPR 2022 · similarity 0.89)
- Learnable Graph Matching: Incorporating Graph Partitioning With Deep Feature Learning for Multiple Object Tracking (CVPR 2021 · similarity 0.89)
- Self-Supervised Multi-Object Tracking with Path Consistency (CVPR 2024 · similarity 0.86)
- Multi-Scene Generalized Trajectory Global Graph Solver with Composite Nodes for Multiple Object Tracking (AAAI 2024 · similarity 0.85)
- Multiple Object Tracking as ID Prediction (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.