Research map / Detection and segmentation / Tracking
MOTRv2: Bootstrapping End-to-End Multi-Object Tracking by Pretrained Object Detectors
Read the paperCheck 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 |
|---|---|---|---|
| Use YOLOX proposals as proposal queries (ablation) | MOTRv2 · DanceTrack validation | HOTA: 57.1 -> 60.7 | Helped |
| Add CrowdHuman joint training without YOLOX proposals (ablation) | MOTRv2 · DanceTrack validation | AssA: 49.5 -> 43.9 | Hurt |
| Combine CrowdHuman joint training with YOLOX proposals (ablation) | MOTRv2 · DanceTrack validation | HOTA: 60.7 -> 63.7 | Helped |
| Propagate full boxes instead of center points (ablation) | MOTRv2 · DanceTrack validation | AssA: 49.2 -> 53.1 | Helped |
| Replace learnable positional embeddings with sine-cosine anchor encoding (ablation) | MOTRv2 · DanceTrack validation | HOTA: 63.8 -> 63.7 | No clear effect |
| Encode proposal confidence scores (ablation) | MOTRv2 · DanceTrack validation | HOTA: 63.0 -> 63.7 | Helped |
| Use sine-cosine encoding rather than linear projection for confidence scores (ablation) | MOTRv2 · DanceTrack validation | AssA: 52.1 -> 53.2 | Helped |
| Use query denoising with noise scales 0.4 and 0.4 (ablation) | MOTRv2 · DanceTrack validation | AssA: 53.2 -> 51.5 | Hurt |
| Use query denoising with noise scales 0.1 and 0.05 (ablation) | MOTRv2 · DanceTrack validation | HOTA: 63.7 -> 64.5 | Helped |
| Replace track query anchors with matched YOLOX proposal boxes (ablation) | MOTRv2 · MOT17 valhalf | MOTA: 75.9 -> 84.3 | Helped |
| Remove MOTR predictions unmatched to YOLOX boxes (ablation) | MOTRv2 · MOT17 valhalf | MOTA: 75.9 -> 83.0 | Helped |
| Align predictions and anchors, and remove unmatched predictions (ablation) | MOTRv2 · MOT17 valhalf | MOTA: 75.9 -> 86.8 | Helped |
Closest papers by meaning
- Multiple Object Tracking as ID Prediction (CVPR 2025 · similarity 0.88)
- More Than Meets the Eye: Enhancing Multi-Object Tracking Even with Prolonged Occlusions (ICML 2025 · similarity 0.88)
- OVTR: End-to-End Open-Vocabulary Multiple Object Tracking with Transformer (ICLR 2025 · similarity 0.87)
- S2-Track: A Simple yet Strong Approach for End-to-End 3D Multi-Object Tracking (ICML 2025 · similarity 0.87)
- Focusing on Tracks for Online Multi-Object Tracking (CVPR 2025 · similarity 0.86)
- DiffusionTrack: Diffusion Model for Multi-Object Tracking (AAAI 2024 · 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.