Research map / Detection and segmentation / Tracking
Learning Multi-Object Tracking and Segmentation From Automatic Annotations
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 |
|---|---|---|---|
| Train MOTSNet on automatically generated KITTI Synth data rather than only KITTI MOTS training images (ablation) | MOTSNet · KITTI MOTS val (Car), sMOTSA | sMOTSA: 72.6 -> 77.6 | Helped |
| Add Mapillary Vistas pre-training to ImageNet initialization (ablation) | MOTSNet · KITTI MOTS val (Ped), sMOTSA | sMOTSA: 49.1 -> 54.5 | Helped |
| Add mask-pooling in the Tracking Head instead of whole-box average pooling (ablation) | MOTSNet · KITTI MOTS val (Car), sMOTSA | sMOTSA: 73.7 -> 77.6 | Helped |
| Add mask-pooling in the Tracking Head instead of whole-box average pooling (ablation) | MOTSNet · KITTI MOTS val (Ped), sMOTSA | sMOTSA: 46.4 -> 49.1 | Helped |
| Add a learned Tracking Head and its loss to mask-based pooling (ablation) | MOTSNet · KITTI MOTS val (Car), sMOTSA | sMOTSA: 76.4 -> 77.6 | Helped |
| Add a learned Tracking Head and its loss to mask-based pooling (ablation) | MOTSNet · KITTI MOTS val (Ped), sMOTSA | sMOTSA: 44.0 -> 49.1 | Helped |
| Add KITTI Synth pre-training to ImageNet and Mapillary Vistas pre-training (ablation) | MOTSNet · KITTI MOTS val (Car), sMOTSA | sMOTSA: 74.9 -> 78.1 | Helped |
| Add COCO pre-training to ImageNet initialization (ablation) | MOTSNet · MOTSChallenge, sMOTSA | sMOTSA: 41.8 -> 56.8 | Helped |
| Pool under the detected mask rather than over the whole box (ablation) | MOTSNet on BDD100k · BDD100k val, MOTA | MOTA: 56.9 -> 58.2 | Helped |
| Add temporal proximity to embedding-distance association (ablation) | MOTSNet track extraction · KITTI MOTS (Ped), sMOTSA | sMOTSA: 54.4 -> 54.6 | Helped |
| Add temporal proximity to embedding-distance association (ablation) | MOTSNet track extraction · KITTI MOTS (Car), sMOTSA | sMOTSA: 78.2 -> 78.1 | Hurt |
| Use embedding distance instead of sIoU alone for track association (ablation) | MOTSNet track extraction · KITTI MOTS (Car), sMOTSA | sMOTSA: 76.7 -> 78.2 | Helped |
Closest papers by meaning
- MOTS: Multi-Object Tracking and Segmentation (CVPR 2019 · similarity 0.94)
- SAM2MOT: A Novel Paradigm of Multi-Object Tracking by Segmentation (AAAI 2026 · similarity 0.90)
- Multiple Object Tracking as ID Prediction (CVPR 2025 · similarity 0.87)
- More Than Meets the Eye: Enhancing Multi-Object Tracking Even with Prolonged Occlusions (ICML 2025 · similarity 0.87)
- Focusing on Tracks for Online Multi-Object Tracking (CVPR 2025 · similarity 0.87)
- OVT-B: A New Large-Scale Benchmark for Open-Vocabulary Multi-Object Tracking (NeurIPS 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.