Research map / Detection and segmentation / Weakly supervised detection and segmentation
Object-Centric Multiple Object Tracking
Zixu Zhao, Jiaze Wang, Max Horn, Yizhuo Ding, Tong He, Zechen Bai, Dominik Zietlow, Carl-Johann Simon-Gabriel, Bing Shuai, Zhuowen Tu, Thomas Brox, Bernt Schiele and 4 more
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 |
|---|---|---|---|
| add object memory rollout instead of last tracks as index query (ablation) | OC-MOT · CATER | IDF1: 77.2% -> 88.6% | Improved |
| use two MHA modules instead of dot-product similarity (ablation) | OC-MOT index-merge · CATER | IDF1: 72.0% -> 88.6% | Improved |
| use two MHA instead of one shared MHA (ablation) | OC-MOT index-merge · CATER | IDF1: 86.2% -> 88.6% | Improved |
| increase memory length Tmax from 6 to 32 (consecutive) (ablation) | OC-MOT · CATER | IDF1: 82.9% -> 88.4% | Improved |
| use slow-fast sequence sampling instead of consecutive at Tmax=6 (ablation) | OC-MOT · CATER | IDF1: 82.9% -> 88.6% | Improved |
| add 6.25% sparse mask labels with DETR loss to DINOSAUR grouping (ablation) | DINOSAUR · FISHBOWL | FG-ARI | Improved |
| double number of annotated frames from 4 to 8 (ablation) | DINOSAUR+DETR · FISHBOWL | FG-ARI: +1.6% | No clear effect |
| OC-MOT vs SAVi using object-centric association | SAVi · FISHBOWL | IDF1: 46.9% -> 77.9% | Improved |
| reconstruct optical flow instead of RGB in SAVi | SAVi · FISHBOWL | IDF1: 46.9% -> 53.2% | Mixed |
| unsupervised OC-MOT versus supervised MOTR on FISHBOWL | OC-MOT · FISHBOWL | IDF1: 81.6% -> 77.9% | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SAVi — SAVi++: Towards End-to-End Object-Centric Learning from Real-World Videos (NeurIPS 2022) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Multiple Object Tracking as ID Prediction (CVPR 2025 · similarity 0.90)
- MOTS: Multi-Object Tracking and Segmentation (CVPR 2019 · similarity 0.89)
- Benchmarking Unsupervised Object Representations for Video Sequences (JMLR 2021 · similarity 0.89)
- From Detection to Association: Learning Discriminative Object Embeddings for Multi-Object Tracking (CVPR 2026 · similarity 0.89)
- COVTrack++: Learning Open-Vocabulary Multi-Object Tracking from Continuous Videos via a Synergistic Paradigm (TPAMI 2026 · similarity 0.89)
- TransCenter: Transformers With Dense Representations for Multiple-Object Tracking (TPAMI 2022 · similarity 0.89)
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.
