Research map / Video / Video segmentation and tracking
Ego-Surfing: Person Localization in First-Person Videos Using Ego-Motion Signatures
Ryo Yonetani, Kris Kitani, Yoichi Sato
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 data-driven generic targetness to correlation-based targetness (ablation) | Proposed · Ego-Surfing dataset | AUC: 0.80 -> 0.84 | Improved |
| add correlation-based targetness to generic-only targetness (ablation) | Proposed · Ego-Surfing dataset | AUC: 0.81 -> 0.84 | Improved |
| add correlation-based targetness to generic-only targetness (ablation) | Proposed · CMU dataset | AUC: 0.82 -> 0.83 | No clear effect |
| add generic targetness to correlation-only (ablation) | Proposed · CMU dataset | AUC: 0.78 -> 0.83 | Improved |
| use trajectory candidates with generic targetness instead of supervoxel candidates (YKS) | Proposed · Ego-Surfing dataset | AUC: 0.79 -> 0.84 | Improved |
| efficient search ranking by correlation upperbound instead of generic targetness (ablation) | Proposed-ES · Ego-Surfing dataset | AUC: 0.76 -> 0.85 | Improved |
| efficient search using correlation upperbound with P=25 (ablation) | Proposed-ES vs Proposed · Ego-Surfing dataset | AUC: 0.84 -> 0.85 | No clear effect |
| reduce candidates evaluated below P=25 (ablation) | Proposed-ES · Ego-Surfing and CMU datasets | AUC and AP | Hurt |
| efficient search (P=25) instead of evaluating all candidates | Proposed-ES · Ego-Surfing dataset | computation time: 69 percent per-candidate, 88 percent per-pixel reduction | Improved |
| use symmetric affinity A(Vp,Vq) instead of asymmetric (ablation) | Proposed · Target video retrieval | mean R-precision: 0.50 -> 0.57 | Improved |
| use P(V|a) instead of P(a|V) targetness in affinity (ablation) | Proposed · Target video retrieval | mean R-precision: 0.49 -> 0.57 | Improved |
| use symmetric affinity instead of asymmetric (ablation) | Proposed · Social group clustering | F-measure: 0.60 -> 0.84 | Improved |
Closest papers by meaning
- Future Person Localization in First-Person Videos (CVPR 2018 · similarity 0.90)
- EgoPrivacy: What Your First-Person Camera Says About You? (ICML 2025 · similarity 0.88)
- Ego-Only: Egocentric Action Detection without Exocentric Transferring (ICCV 2023 · similarity 0.87)
- EgoSim: An Egocentric Multi-view Simulator and Real Dataset for Body-worn Cameras during Motion and Activity (NeurIPS 2024 · similarity 0.86)
- Estimating Ego-Body Pose from Doubly Sparse Egocentric Video Data (NeurIPS 2024 · similarity 0.86)
- Instance Tracking in 3D Scenes from Egocentric Videos (CVPR 2024 · 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.
