Research map / Video / Video-language understanding
A Multi-modal Global Instance Tracking Benchmark (MGIT): Better Locating Target in Complex Spatio-temporal and Causal Relationship
Shiyu Hu, Dailing Zhang, Meiqi Wu, Xiaokun Feng, Xuchen Li, Xin Zhao, Kaiqi Huang
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 story-level description instead of first-action description (mechanism D vs A) (ablation) | SNLT · MGIT | SR: 0.036 -> 0.040 | Improved |
| use story-level description instead of first-action description (mechanism D vs A) (ablation) | VLT_SCAR · MGIT | SR: 0.177 -> 0.184 | Improved |
| use story-level description instead of first-action description (mechanism D vs A) (ablation) | VLT_TT · MGIT | SR: 0.474 -> 0.480 | Improved |
| use story-level description instead of first-action description (mechanism D vs A) (ablation) | JointNLT · MGIT | SR: 0.603 -> 0.600 | Hurt |
| use activity instead of action granularity (ablation) | VLT_TT · MGIT | SR: 0.468 -> 0.485 | Improved |
| use activity instead of action granularity (ablation) | VLT_SCAR · MGIT | SR: 0.167 -> 0.180 | Improved |
| use activity instead of action granularity (ablation) | JointNLT · MGIT | SR: 0.610 -> 0.605 | Hurt |
| use ambiguous descriptions (LaSOTSub) rather than unambiguous (LaSOTNLC) (ablation) | VLT_TT · LaSOT subsets | SR: 0.662 -> 0.660 | Hurt |
| use ambiguous descriptions (LaSOTSub) rather than unambiguous (LaSOTNLC) (ablation) | JointNLT · LaSOT subsets | SR: 0.651 -> 0.583 | Hurt |
| evaluate trackers on MGIT instead of LaSOT | VLT_TT · MGIT vs LaSOT | SR: 0.670 -> 0.474 | Hurt |
| evaluate trackers on MGIT instead of LaSOT | JointNLT · MGIT vs LaSOT | SR: 0.607 -> 0.603 | Hurt |
| transformer architecture instead of Siamese network, BBox only | MixFormer vs SiamCAR · MGIT | SR: 0.183 -> 0.629 | Improved |
Closest papers by meaning
- Beyond MOT: Semantic Multi-Object Tracking (ECCV 2024 · similarity 0.91)
- Interactive Tracking: A Human-in-the-Loop Paradigm with Memory-Augmented Adaptation (CVPR 2026 · similarity 0.90)
- Single-Model and Any-Modality for Video Object Tracking (CVPR 2024 · similarity 0.89)
- UniSOT: A Unified Framework for Multi-Modality Single Object Tracking (TPAMI 2025 · similarity 0.89)
- VastTrack: Vast Category Visual Object Tracking (NeurIPS 2024 · similarity 0.88)
- MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models (CVPR 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.
