Research map / Video / Video-language understanding
Fine-grained Spatiotemporal Grounding on Egocentric Videos
Shuo Liang, Yiwu Zhong, Zi-Yuan Hu, Yeyao Tao, Liwei Wang
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 fine-tuning on EgoMask-Train | VideoLISA-3.8B · EgoMask-Short | IoU_gold_pred: 17.85 -> 23.36 | Improved |
| add fine-tuning on EgoMask-Train | VideoLISA-3.8B · EgoMask-Medium | IoU_gold_pred: 6.48 -> 9.98 | Improved |
| add fine-tuning on EgoMask-Train | VideoLISA-3.8B · EgoMask-Long | IoU_gold_pred: 5.15 -> 7.16 | Improved |
| add fine-tuning on EgoMask-Train | Sa2VA-4B · EgoMask-Short | IoU_gold_pred: 29.00 -> 30.97 | Improved |
| add fine-tuning on EgoMask-Train | Sa2VA-4B · EgoMask-Long | IoU_gold_pred: 8.11 -> 8.24 | No clear effect |
| add fine-tuning on EgoMask-Train | Sa2VA-4B · Mevis (exocentric) | IoU_gold_pred: 50.01 -> 55.55 | Improved |
| add fine-tuning on EgoMask-Train | VideoLISA-3.8B · Ref-Davis (exocentric) | IoU_gold_pred: 65.82 -> 65.60 | No clear effect |
| add fine-tuning on EgoMask-Train | VideoLISA-3.8B · ReasonVOS (exocentric) | IoU_gold_pred: 42.41 -> 44.18 | Improved |
| use highest-confidence GroundingDINO box to initialize SAM2 (ablation) | Grounded-SAM2 · EgoMask-Short | IoU_gold_pred: 40.42 -> 49.95 | Improved |
| use highest-confidence GroundingDINO box to initialize SAM2 (ablation) | Grounded-SAM2 · EgoMask-Long | IoU_gold_pred: 11.65 -> 24.80 | Improved |
| valid target in first five key frames for SAM2 initialization (ablation) | Sa2VA-26B · EgoMask-Medium | IoU_gold_pred: 0.86 -> 34.15 | Improved |
| scale Sa2VA from 4B to 26B | Sa2VA · EgoMask-Long | IoU_gold_pred: 8.11 -> 12.96 | Improved |
Closest papers by meaning
- Object-Shot Enhanced Grounding Network for Egocentric Video (CVPR 2025 · similarity 0.92)
- Grounded-Exo2Ego: Structured Semantic Grounding for Robust Exocentric-to-Egocentric Video Generation (NeurIPS 2026 · similarity 0.89)
- EgoThinker: Unveiling Egocentric Reasoning with Spatio-Temporal CoT (NeurIPS 2025 · similarity 0.89)
- Learning to Segment Referred Objects from Narrated Egocentric Videos (CVPR 2024 · similarity 0.88)
- VideoGEM: Training-free Action Grounding in Videos (CVPR 2025 · similarity 0.88)
- Action Scene Graphs for Long-Form Understanding of Egocentric Videos (CVPR 2024 · 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.
