Research map / Vision-language and multimodal / Cross-modal learning
RAGTrack: Language-aware RGBT Tracking with Retrieval-Augmented Generation
Hao Li, Yuhao Wang, Wenning Hao, Pingping Zhang, Dong Wang, Huchuan Lu
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 text-free context-aware reasoning module (CRM*) to baseline (ablation) | Baseline tracker · RGBT234 | MPR: 87.9 -> 89.1 | Improved |
| add multi-modal transformer encoder with text on top of CRM* (ablation) | Baseline + CRM* · RGBT234 | MPR: 89.1 -> 91.1 | Improved |
| use language-based CRM instead of text-free CRM* (ablation) | MTE + CRM · RGBT234 | MPR: 91.1 -> 91.8 | Improved |
| add Adaptive Token Fusion (ATF) (ablation) | MTE + CRM · RGBT234 | MPR: 91.8 -> 93.8 | Improved |
| apply ATF at four layers (6,12,18,24) instead of one (ablation) | RAGTrack · RGBT234 | MSR: 67.9 -> 69.5 | Improved |
| use ATF instead of TBSI, BSI or DFM fusion (ablation) | RAGTrack · RGBT234 | MSR: 68.2 -> 69.5 vs best alternative (BSI) | Improved |
| use proposed augmentation instead of addition, Transformer or Mamba (ablation) | CRM · RGBT234 | MSR: 69.0 -> 69.5 vs best alternative (Transformer) | Improved |
| increase reasoning tokens from 1 to 4 (ablation) | CRM · RGBT234 | MPR: 93.8 -> 93.2 | Hurt |
| add 2 learnable tokens to the text prefix instead of 0 (ablation) | MTE · RGBT234 | MPR: 92.8 -> 93.8 | Improved |
| increase learnable tokens from 2 to 4 (ablation) | MTE · RGBT234 | MSR: 69.5 -> 68.7 | Hurt |
| use all four attention scores for token selection instead of one (ablation) | ATF · RGBT234 | MPR: 92.0 -> 93.8 | Improved |
| set retention ratio to 85% instead of lower or higher values (ablation) | ATF · RGBT234 | MPR | Improved |
Closest papers by meaning
- RGB-T Tracking With Template-Bridged Search Interaction and Target-Preserved Template Updating (TPAMI 2024 · similarity 0.90)
- Temporal Adaptive RGBT Tracking with Modality Prompt (AAAI 2024 · similarity 0.89)
- Single-Model and Any-Modality for Video Object Tracking (CVPR 2024 · similarity 0.87)
- Cross-modulated Attention Transformer for RGBT Tracking (AAAI 2025 · similarity 0.87)
- CSTrack: Enhancing RGB-X Tracking via Compact Spatiotemporal Features (ICML 2025 · similarity 0.87)
- RGBT Tracking via All-layer Multimodal Interactions with Progressive Fusion Mamba (AAAI 2025 · 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.
