Research map / Detection and segmentation / Tracking
OVTrack: Open-Vocabulary Multiple Object Tracking
Siyuan Li, Tobias Fischer, Lei Ke, Henghui Ding, Martin Danelljan, Fisher Yu
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 OVTrack instead of video-trained closed-set trackers | TAO tracking · TAO validation, base and novel classes | TETA: 35.5 vs 30.3; 27.8 vs 25.7 | Improved |
| Use OVTrack instead of video-trained closed-set trackers | TAO tracking · TAO test, base and novel classes | TETA: 32.6 vs 29.2; 24.1 vs 21.7 | Improved |
| Distill image embeddings in addition to text embeddings (ablation) | OVTrack · TAO validation | ClsA: 15.6 -> 18.1 | Improved |
| Add image distillation to text distillation (ablation) | OVTrack · TAO validation | LocA: 50.5 -> 49.3 | Hurt |
| Add DDPM hallucination to standard augmentations (ablation) | TETer-SwinT · TAO validation | AssocA: 30.6 -> 33.0 | Improved |
| Add heavy augmentations to DDPM and standard augmentations (ablation) | TETer-SwinT · TAO validation | AssocA: 33.0 -> 35.5 | Improved |
| Add DDPM hallucination to standard augmentations (ablation) | OVTrack · TAO validation | AssocA: 31.1 -> 32.9 | Improved |
| Add heavy augmentations to DDPM and standard augmentations (ablation) | OVTrack · TAO validation | AssocA: 32.9 -> 35.7 | Improved |
| Use DDPM without language prompts or geometric transformation (ablation) | OVTrack · TAO validation | AssocA: 31.1 -> 30.6 | Hurt |
| Add geometric transformation before DDPM generation (ablation) | OVTrack · TAO validation | AssocA: 31.1 -> 32.4 | Improved |
| Add language prompting to DDPM with geometric transformation (ablation) | OVTrack · TAO validation | AssocA: 32.4 -> 32.9 | Improved |
| Transfer OVTrack trained on LVIS rather than target-dataset training | BDD100K tracking · BDD100K validation | TETA: 42.5 vs 36.1 | Improved |
Later papers that compared with this method
11 later accepted papers report OVTrack as a baseline in their results tables; 9 of them report a better number on most of the results they share with it.
- SOVTrack: Open-Vocabulary Multi-Object Tracking with Self-Supervised Pseudo Labeling and Feature Distillation (ECCV 2026 · better on 4 of 4 shared results)
- SAM2-OV: A Novel Detection-Only Tuning Paradigm for Open-Vocabulary Multi-Object Tracking (AAAI 2026 · better on 0 of 1 shared result)
- COVTrack++: Learning Open-Vocabulary Multi-Object Tracking from Continuous Videos via a Synergistic Paradigm (TPAMI 2026 · better on 1 of 1 shared result)
- OVTR: End-to-End Open-Vocabulary Multiple Object Tracking with Transformer (ICLR 2025 · better on 6 of 8 shared results)
- COVTrack: Continuous Open-Vocabulary Tracking via Adaptive Multi-Cue Fusion (ICCV 2025 · better on 8 of 8 shared results)
- Lattice Boltzmann Model for Learning Real-World Pixel Dynamicity (NeurIPS 2025 · better on 1 of 1 shared result)
- VOVTrack: Exploring the Potentiality in Raw Videos for Open-Vocabulary Multi-Object Tracking (ICCV 2025 · better on 1 of 1 shared result)
- OVT-B: A New Large-Scale Benchmark for Open-Vocabulary Multi-Object Tracking (NeurIPS 2024 · better on 6 of 8 shared results)
3 more not listed.
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
- OVT-B: A New Large-Scale Benchmark for Open-Vocabulary Multi-Object Tracking (NeurIPS 2024 · similarity 0.95)
- VOVTrack: Exploring the Potentiality in Raw Videos for Open-Vocabulary Multi-Object Tracking (ICCV 2025 · similarity 0.94)
- OVTR: End-to-End Open-Vocabulary Multiple Object Tracking with Transformer (ICLR 2025 · similarity 0.93)
- VastTrack: Vast Category Visual Object Tracking (NeurIPS 2024 · similarity 0.91)
- DiffusionTrack: Diffusion Model for Multi-Object Tracking (AAAI 2024 · similarity 0.90)
- Multiple Object Tracking as ID Prediction (CVPR 2025 · similarity 0.90)
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.
