Research map / Detection and segmentation / Tracking
SOTVerse: A User-Defined Task Space of Single Object Tracking
Shiyu Hu, Xin Zhao, Kaiqi Huang
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 |
|---|---|---|---|
| evaluate only challenging frames instead of all frames | KeepTrack · OTB2015 | success rate: 90% -> 46.8% | Hurt |
| relax task constraints from short-term to global instance tracking | 20 trackers · SOTVerse normal space | tracking performance | Hurt |
| move from OTB2015 to VideoCube (shot-cut global instance tracking) | KeepTrack · OTB2015 vs VideoCube | precision score: 0.927 -> 0.291 | Hurt |
| move from OTB2015 to VideoCube | MixFormer · OTB2015 vs VideoCube | precision score: 0.888 -> 0.402 | Hurt |
| add re-initialization (R-OPE) to OPE evaluation | trackers · SOTVerse normal space | precision, normalized precision, success | Improved |
| add re-initialization (R-OPE) to OPE evaluation | trackers · SOTVerse normal space | challenging plot success | No clear effect |
| add re-initialization for short-term tracking | trackers · short-term tracking (OTB, VOT, GOT-10k) | score difference OPE vs R-OPE | No clear effect |
| add re-initialization for long-term and global instance tracking | trackers · long-term tracking, GIT | longest tracked sub-sequence | Improved |
| weight challenging-space results by sequence length | top-3 trackers · challenging spaces | precision, normalized precision, success | Hurt |
| use end-to-end transformer mixed attention (MixFormer) instead of other architectures | MixFormer · SOTVerse mean precision | precision score: 0.661 -> 0.704 | Improved |
| use deep tracker instead of correlation filter KCF | KeepTrack · SOTVerse mean precision | precision score: 0.219 -> 0.637 | Improved |
Closest papers by meaning
- TAO: A Large-Scale Benchmark for Tracking Any Object (ECCV 2020 · similarity 0.90)
- M3SOT: Multi-Frame, Multi-Field, Multi-Space 3D Single Object Tracking (AAAI 2024 · similarity 0.89)
- SUTrack: Towards Simple and Unified Single Object Tracking (AAAI 2025 · similarity 0.89)
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · similarity 0.87)
- OVT-B: A New Large-Scale Benchmark for Open-Vocabulary Multi-Object Tracking (NeurIPS 2024 · similarity 0.87)
- VastTrack: Vast Category Visual Object Tracking (NeurIPS 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.
