Research map / Vision-language and multimodal / Human and embodied modeling
Generalizing Multiple Object Tracking to Unseen Domains by Introducing Natural Language Representation
En Yu, Songtao Liu, Zhuoling Li, Jinrong Yang, Zeming Li, Shoudong Han, Wenbing Tao
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 VCP and VLM pseudo textual descriptions to MOTR | MOTR · MOT20 validation (cross-domain) | HOTA: 43.7 -> 46.8 | Improved |
| add VCP and VLM pseudo textual descriptions to MOTR | MOTR · MOT20 validation (cross-domain) | AssA: 41.7 -> 45.4 | Improved |
| add VCP and VLM pseudo textual descriptions to MOTR | MOTR · MOT20 validation (cross-domain) | IDF1: 56.8 -> 61.1 | Improved |
| add VCP and VLM pseudo textual descriptions to MOTR | MOTR · BDD100K pedestrian | HOTA-pedestrian: 32.4 -> 33.7 | Improved |
| use final Transformer encoder output rather than backbone-stage features as VCP context (ablation) | LTrack VCP · MOT20 validation (cross-domain, lite LTrack) | HOTA: 15.8 / 18.3 / 15.5 -> 19.2 | Improved |
| combine textual and visual prompts rather than using either alone (ablation) | LTrack VLM · MOT20 validation (cross-domain, lite LTrack) | HOTA: 16.1 / 17.5 -> 19.2 | Improved |
| add the adapter to VLM (ablation) | LTrack VLM · MOT20 validation (cross-domain, lite LTrack) | HOTA: 16.7 -> 19.2 | Improved |
| use the template 'A photo of' rather than 'a' (ablation) | LTrack textual prompts · MOT20 validation (cross-domain, lite LTrack) | HOTA: 17.0 -> 19.2 | Improved |
| vary text-token representation length; use length 15 rather than lengths that are too small or large (ablation) | LTrack textual prompts · MOT20 validation (cross-domain) | HOTA | Mixed |
| introduce natural-language-derived pseudo textual descriptions into the tracker | LTrack · MOT17 validation | HOTA: 57.2 -> 57.5 | Improved |
Closest papers by meaning
- Beyond Explicit Language: Plug-and-Play Visual-to-Linguistic Modeling Toward General Object Tracking (CVPR 2026 · similarity 0.91)
- ChatTracker: Enhancing Visual Tracking via LLM-Driven Iterative Description Refinement (TPAMI 2026 · similarity 0.91)
- ChatTracker: Enhancing Visual Tracking Performance via Chatting with Multimodal Large Language Model (NeurIPS 2024 · similarity 0.90)
- Towards Generalizable Multi-Object Tracking (CVPR 2024 · similarity 0.89)
- Unifying Visual and Vision-Language Tracking via Contrastive Learning (AAAI 2024 · similarity 0.88)
- OVTR: End-to-End Open-Vocabulary Multiple Object Tracking with Transformer (ICLR 2025 · similarity 0.88)
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.
