Research map / 3D vision / Human pose estimation
Future Person Localization in First-Person Videos
Takuma Yagi, Karttikeya Mangalam, Ryo Yonetani, 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 person scale to location inputs (ablation) | Future person localization model · FPL Dataset | FDE (pixels), Average: 88.16 -> 81.86 | Improved |
| Add ego-motion to location-scale inputs (ablation) | Future person localization model · FPL Dataset | FDE (pixels), Average: 81.86 -> 79.09 | Improved |
| Add pose to location-scale inputs (ablation) | Future person localization model · FPL Dataset | FDE (pixels), Average: 81.86 -> 80.57 | Improved |
| Combine ego-motion and pose with location-scale cues (ablation) | Future person localization model · FPL Dataset | FDE (pixels), Average: 80.57 -> 77.26 | Improved |
| Add ego-motion to location-scale inputs (ablation) | Future person localization model · First-Person Social Interaction Dataset | FDE (pixels), Average: 127.40 -> 127.28 | No clear effect |
| Add pose to location-scale inputs (ablation) | Future person localization model · First-Person Social Interaction Dataset | FDE (pixels), Average: 127.40 -> 125.78 | Improved |
| Combine ego-motion and pose with location-scale cues (ablation) | Future person localization model · First-Person Social Interaction Dataset | FDE (pixels), Average: 125.78 -> 125.42 | Improved |
| Shorten the input history from 10 frames to 6 frames (ablation) | Future person localization model · FPL Dataset, 10-frame future | FDE (pixels), Average: 77.26 -> 79.77 | Hurt |
| Lengthen the input history from 6 frames to 10 frames (ablation) | Future person localization model · FPL Dataset, 6-frame future | FDE (pixels), Average: 46.16 -> 45.92 | Improved |
| Predict a 20-frame future instead of the standard 10-frame future (ablation) | Future person localization model · FPL Dataset | FDE (pixels), Average: 77.26 -> 124.85 | Hurt |
Closest papers by meaning
- Ego-Surfing: Person Localization in First-Person Videos Using Ego-Motion Signatures (TPAMI 2017 · similarity 0.90)
- Subjects and Their Objects: Localizing Interactees for a Person-Centric View of Importance (IJCV 2016 · similarity 0.86)
- Forecasting Human-Object Interaction: Joint Prediction of Motor Attention and Actions in First Person Video (ECCV 2020 · similarity 0.86)
- Identifying First-Person Camera Wearers in Third-Person Videos (CVPR 2017 · similarity 0.86)
- Actor and Observer: Joint Modeling of First and Third-Person Videos (CVPR 2018 · similarity 0.86)
- Sharingan: A Transformer Architecture for Multi-Person Gaze Following (CVPR 2024 · similarity 0.84)
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.
