Research map / Detection and segmentation / Tracking
Robust Multi-Modality Multi-Object Tracking
Wenwei Zhang, Hui Zhou, Shuyang Sun, Zhe Wang, Jianping Shi, Chen Change Loy
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 |
|---|---|---|---|
| use point cloud instead of image features for data association (ablation) | 2D tracker baseline · KITTI tracking val (PointPillar detections) | MOTA: 74.88 -> 75.70 | Improved |
| ensemble image and point cloud models (ablation) | single-sensor baselines · KITTI tracking val (PointPillar detections) | MOTA: 75.70 -> 77.54 | Improved |
| add attention fusion (module C) in robust fusion (ablation) | single-sensor baselines · KITTI tracking val (PointPillar detections) | MOTA: 77.54 -> 78.18 | Improved |
| use concatenation fusion (A) in robust module vs ensemble (ablation) | ensemble baseline · KITTI tracking val (PointPillar detections) | MOTA: 77.54 -> 77.57 | No clear effect |
| use addition fusion (B) in robust module vs ensemble (ablation) | ensemble baseline · KITTI tracking val (PointPillar detections) | MOTA: 77.54 -> 77.62 | No clear effect |
| use robust fusion module instead of normal fusion C (ablation) | fusion module C · KITTI tracking val (PointPillar detections) | MOTA: 77.62 -> 78.18 | Improved |
| use robust fusion module instead of normal fusion A (ablation) | fusion module A · KITTI tracking val (PointPillar detections) | MOTA: 77.31 -> 77.57 | Improved |
| use absolute subtraction instead of multiplication as correlation (ablation) | 2D tracker · KITTI tracking val (PointPillar detections) | MOTA: 74.88 -> 77.76 | Improved |
| use softmax with addition ranking mechanism instead of none (ablation) | 2D tracker · KITTI tracking val (PointPillar detections) | MOTA: 74.88 -> 77.40 | Improved |
| use softmax with multiplication ranking (ablation) | 2D tracker · KITTI tracking val (PointPillar detections) | MOTA: 74.88 -> 75.08 | No clear effect |
| add absolute subtraction correlation to fusion model (ablation) | fusion C · KITTI tracking val (PointPillar detections) | MOTA: 78.18 -> 79.18 | Improved |
| add softmax addition ranking to fusion model (ablation) | fusion C + abs subtraction · KITTI tracking val (PointPillar detections) | MOTA: 79.18 -> 80.08 | Improved |
Closest papers by meaning
- ModTrack: Sensor-Agnostic Multi-View Tracking via Identity-Informed PHD Filtering with Covariance Propagation (ECCV 2026 · similarity 0.88)
- Collaborating Vision, Depth, and Thermal Signals for Multi-Modal Tracking: Dataset and Algorithm (NeurIPS 2025 · similarity 0.86)
- Single-Model and Any-Modality for Video Object Tracking (CVPR 2024 · similarity 0.85)
- Bi-directional Adapter for Multimodal Tracking (AAAI 2024 · similarity 0.85)
- Layer-Wise Modality Decomposition for Interpretable Multimodal Sensor Fusion (NeurIPS 2025 · similarity 0.84)
- Omnidirectional Multi-Object Tracking (CVPR 2025 · similarity 0.84)
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.
