Research map / Detection and segmentation / Tracking
Eliminating Exposure Bias and Metric Mismatch in Multiple 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 |
|---|---|---|---|
| Build a rich training set iteratively from tracklets encountered during tracking (ablation) | OURS-geom · DukeMTMC training validation | IDF: +3.9 IDF versus a fixed set restricted to tracklets with at most one identity switch | Helped |
| Retain hypotheses using the proposed pruning strategy rather than score/count pruning (ablation) | OURS-geom · DukeMTMC training validation | IDF: +20 IDF versus score/count pruning | Helped |
| Predict per-frame quantities and use them to approximate IDF rather than regress IDF directly (ablation) | OURS-geom · DukeMTMC training validation | IDF: +5.1 IDF | Helped |
| Add bounding-box shift regression (ablation) | OURS-geom · DukeMTMC training validation | IDF: +2.2 IDF | Helped |
| Use the proposed scoring loss rather than a standard classification loss (ablation) | OURS-geom · DukeMTMC training validation | IDF: +32.8 IDF | Helped |
| Balance the training set across IDF values (ablation) | OURS-geom · DukeMTMC training validation | IDF: +4.7 IDF | Helped |
| Use hard-mining during training (ablation) | OURS-geom · DukeMTMC training validation | IDF: +2.5 IDF | Helped |
| Replace greedy final selection with an Integer Program (ablation) | OURS-geom · DukeMTMC training validation | IDF | No clear effect |
| Pre-train the model separately with each type of feature (ablation) | OURS-geom · DukeMTMC training validation | IDF | No clear effect |
| Use a deeper network (ablation) | OURS-geom · DukeMTMC training validation | IDF | No clear effect |
| Add appearance features (ablation) | OURS-geom · DukeMTMC | IDF: 74.6 to 82.5 | Helped |
| Add social features (ablation) | OURS-geom · DukeMTMC | IDF: 67.5 to 74.6 | Helped |
Closest papers by meaning
- Multiple Object Tracking as ID Prediction (CVPR 2025 · similarity 0.89)
- Focusing on Tracks for Online Multi-Object Tracking (CVPR 2025 · similarity 0.87)
- More Than Meets the Eye: Enhancing Multi-Object Tracking Even with Prolonged Occlusions (ICML 2025 · similarity 0.87)
- DiffusionTrack: Diffusion Model for Multi-Object Tracking (AAAI 2024 · similarity 0.87)
- Self-Supervised Multi-Object Tracking with Path Consistency (CVPR 2024 · similarity 0.87)
- Towards Generalizable Multi-Object Tracking (CVPR 2024 · 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.