Research map / Architectures and efficiency / Pruning and efficient tuning
CAT: A Unified Click-and-Track Framework for Realistic Tracking
Yongsheng Yuan, Jie Zhao, 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 JSVR second-stage refinement (ablation) | CATos256 · UAVTrack112 | AUC: 59.8 -> 67.7 | Improved |
| add JSVR second-stage refinement (ablation) | CATos256 · LaSOT | AUC: 66.1 -> 66.6 | Improved |
| use tracking-based refinement instead of JSVR (ablation) | CATos256 · LaSOT | AUC: 66.1 -> 64.9 | Hurt |
| use detection-trained second stage instead of JSVR (ablation) | CATos256 · LaSOT | AUC: 66.3 -> 66.6 | Hurt |
| apply JSVR to SAM1B initial boxes (ablation) | SAM1B · UAV123 | AUC: 51.8 -> 57.8 (+6.0) | Improved |
| apply JSVR to SAM1H initial boxes (ablation) | SAM1H · UAV123 | AUC: 51.9 -> 57.3 (+5.4) | Improved |
| apply JSVR to SAM2B initial boxes (ablation) | SAM2B · LaSOT | AUC: 62.4 -> 62.6 (+0.2) | Improved |
| apply JSVR to SAM2L initial boxes (ablation) | SAM2L · DTB70 | AUC: 60.0 -> 58.7 (-1.3) | Hurt |
| apply JSVR to SAM2L initial boxes (ablation) | SAM2L · UAV123 | AUC: 63.5 -> 63.7 (+0.2) | No clear effect |
| set CTMoE expert dimension to 128 instead of 256 (ablation) | CATos256 · UAVTrack112 | AUC: 66.0 -> 67.7 | Improved |
| set CTMoE expert dimension to 128 instead of 16 (ablation) | CATos256 · LaSOT | AUC: 66.4 -> 66.6 | Improved |
| enlarge click sampling region from 10% to 25% (ablation) | CATos384 · UAVTrack112 | IoU: 78.6 -> 76.3 | Hurt |
Closest papers by meaning
- Interactive Tracking: A Human-in-the-Loop Paradigm with Memory-Augmented Adaptation (CVPR 2026 · similarity 0.88)
- Multi-Timescale Collaborative Tracking (TPAMI 2016 · similarity 0.88)
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · similarity 0.87)
- Focusing on Tracks for Online Multi-Object Tracking (CVPR 2025 · similarity 0.87)
- OneTracker: Unifying Visual Object Tracking with Foundation Models and Efficient Tuning (CVPR 2024 · similarity 0.86)
- Is This Tracker On? A Benchmark Protocol for Dynamic Tracking (NeurIPS 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.
