Research map / Video / Video-language understanding
T2SGrid: Temporal-to-Spatial Gridification for Video Temporal Grounding
Chaohong Guo, Yihan He, Yongwei Nie, Fei Ma, Xuemiao Xu, Chengjiang Long
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 T2SGrid to a model without temporal encoding | Qwen2-VL-7B · Charades-STA | mIoU: 7.9 -> 44.3 | Improved |
| add T2SGrid to a model with textual timestamps | Qwen3-VL-8B · ActivityNet | R@0.5: 26.2 -> 26.8 | No clear effect |
| add T2SGrid | GPT-4o · Charades-STA | mIoU: 35.4 -> 36.9 | Improved |
| add T2SGrid to an image-only model | LLaVA-OneVision-1.5-8B · Charades-STA | mIoU: 14.5 -> 28.8 | Improved |
| fine-tune with T2SGrid-formatted data instead of standard FT | Qwen2-VL-7B · ActivityNet | mIoU: 37.7 -> 46.7 | Improved |
| fine-tune with T2SGrid-formatted data instead of standard FT | Qwen2-VL-7B · Charades-STA | mIoU: 50.4 -> 53.2 | Improved |
| add T2SGrid to base model on video QA | Qwen2-VL-7B · MVBench | All: 51.7 -> 58.3 | Improved |
| add composite text timestamps (ablation) | Qwen2-VL-7B · Charades-STA | mIoU: 7.9 -> 32.9 | Improved |
| add sliding window (ablation) | Qwen2-VL-7B + ComTextNum · Charades-STA | mIoU: 32.9 -> 36.5 | Improved |
| add grid layout (ablation) | Qwen2-VL-7B + ComTextNum + sliding window · Charades-STA | mIoU: 36.5 -> 44.3 | Improved |
| use grid instead of per-frame visual numbering (no overlap) (ablation) | Qwen2-VL-7B · Charades-STA | mIoU: 38.5 -> 41.2 | Improved |
| add window overlap (stride 12 to 7) (ablation) | T2SGrid g43 · Charades-STA | mIoU: 41.2 -> 44.3 | Improved |
Closest papers by meaning
- Towards Long-Form Spatio-Temporal Video Grounding (ECCV 2026 · similarity 0.90)
- GroundVTS: Visual Token Sampling in Multimodal Large Language Models for Video Temporal Grounding (CVPR 2026 · similarity 0.90)
- ScanFocus: A Coarse-to-Fine Framework for Spatio-Temporal Video Grounding (ECCV 2026 · similarity 0.90)
- VTG-LLM: Integrating Timestamp Knowledge into Video LLMs for Enhanced Video Temporal Grounding (AAAI 2025 · similarity 0.90)
- Time-R1: Post-Training Large Vision Language Model for Temporal Video Grounding (NeurIPS 2025 · similarity 0.89)
- Universal Video Temporal Grounding with Generative Multi-modal Large Language Models (NeurIPS 2025 · similarity 0.89)
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.
