Research map / Vision-language and multimodal / Human and embodied modeling
Learning to Track Instance from Single Nature Language Description
Yaozong Zheng, Bineng Zhong, Qihua Liang, Shuimu Zeng, Haiying Xia, Shuxiang Song
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 dynamic token aggregation module (ablation) | SVLTrack-L256 · LaSOT | AUC: 51.4 -> 52.5 | Improved |
| add weak-to-strong consistency framework (ablation) | SVLTrack-L256 · LaSOT | AUC: 51.6 -> 52.5 | Improved |
| add denoising training strategy (ablation) | SVLTrack-L256 · LaSOT | AUC: 52.0 -> 52.5 | Improved |
| increase search tokens from 4 to 8 (ablation) | SVLTrack-L256 · LaSOT | AUC: 51.9 -> 52.5 | Improved |
| increase search tokens from 8 to 16 (ablation) | SVLTrack-L256 · LaSOT | AUC: 52.5 -> 52.1 | Hurt |
| use Euclidean distance instead of cross-entropy to rank noisy samples (ablation) | SVLTrack-L256 · LaSOT | AUC: 47.9 -> 52.5 | Improved |
| use APE instead of LISA to generate pseudo boxes (ablation) | SVLTrack-L256 · LaSOT | AUC: 48.4 -> 52.5 | Improved |
| set denoising ratio to 20% (ablation) | SVLTrack · LaSOT | AUC: 20% | Improved |
| scale input resolution from 256 to 384 | SVLTrack-L · LaSOT | AUC: 52.5 -> 53.9 | Improved |
| use language-only self-supervised SVLTrack-L256 instead of ATTracker with box and language | ATTracker · TNL2K | AUC: 40.6 -> 42.5 | Improved |
| use SVLTrack-V256 instead of Diff-Tracker | Diff-Tracker · LaSOT | AUC: 48.6 -> 65.1 | Improved |
| compare SVLTrack-L256 with fully supervised language-init JointNLT | JointNLT · LaSOT | AUC: 56.9 -> 52.5 | Hurt |
Closest papers by meaning
- ChatTracker: Enhancing Visual Tracking via LLM-Driven Iterative Description Refinement (TPAMI 2026 · similarity 0.92)
- ChatTracker: Enhancing Visual Tracking Performance via Chatting with Multimodal Large Language Model (NeurIPS 2024 · similarity 0.89)
- Unifying Visual and Vision-Language Tracking via Contrastive Learning (AAAI 2024 · similarity 0.89)
- MemVLT: Vision-Language Tracking with Adaptive Memory-based Prompts (NeurIPS 2024 · similarity 0.87)
- MVLM: Template-Free Tracking via Vision-Language Margin Confidence and Memory-Gated Tracking (CVPR 2026 · similarity 0.86)
- Mono3DVLT: Monocular-Video-Based 3D Visual Language Tracking (CVPR 2025 · 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.
