Research map / 3D vision / Human pose estimation
Learning View-Invariant Features for Person Identification in Temporally Synchronized Videos Taken by Wearable Cameras
Kang Zheng, Xiaochuan Fan, Yuewei Lin, Hao Guo, Hongkai Yu, Dazhou Guo, Song 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 motion compensation (ablation) | Proposed (optical flow only) · SEQ 1 | Rank-1 matching rate (%): 72.28 -> 79.82 | Improved |
| add motion compensation (ablation) | Proposed (optical flow only) · SEQ 2 | Rank-1 matching rate (%): 75.68 -> 76.36 | Improved |
| add motion compensation (ablation) | Proposed (optical flow only) · SYN | Rank-1 matching rate (%): 71.06 -> 72.21 | Improved |
| add motion compensation (ablation) | Proposed (optical flow only) · SEQ 1 | Rank-20 matching rate (%): 98.25 -> 97.54 | Hurt |
| pretrain on synthetic Mocap optical flow and 3D skeleton (video+flow+skel.) before fine-tuning (ablation) | Triplet Network · SEQ 1, SEQ 2, SYN | CMC | Improved |
| pretrain on synthetic optical flow only, without skeleton stream (ablation) | Triplet Network · SEQ 1, SEQ 2, SYN | CMC | No clear effect |
| add Mocap synthetic pretraining (ablation) | Triplet Network · SEQ 1 -> SEQ 2 (cross-dataset) | Rank-1 CMC (%): 4.55 -> 11.36 | Improved |
| add Mocap synthetic pretraining (ablation) | Triplet Network · SEQ 1 -> SEQ 2 (cross-dataset) | Rank-20 CMC (%): 50.00 -> 63.64 | Improved |
| combine TN scores with RFA scores (lambda=0.5) | Proposed · SEQ 1 | Rank-1 CMC (%): 79.82 -> 87.02 | Improved |
| combine TN scores with RFA scores (lambda=0.5) | Proposed · SYN | Rank-1 CMC (%): 72.21 -> 82.12 | Improved |
| combine TN scores with RFA scores (lambda=0.5) | Proposed · SEQ 2 | Rank-1 CMC (%): 76.36 -> 82.05 | Improved |
| lower RFA weight lambda from 1.0 to 0.3 | Proposed+RFA · SYN | Rank-1 matching rate (%): 76.92 -> 83.46 | Improved |
Closest papers by meaning
- Cross-View Person Identification by Matching Human Poses Estimated With Confidence on Each Body Joint (AAAI 2018 · similarity 0.90)
- Joint Person Segmentation and Identification in Synchronized First- and Third-person Videos (ECCV 2018 · similarity 0.85)
- Learning Modal-Invariant and Temporal-Memory for Video-Based Visible-Infrared Person Re-Identification (CVPR 2022 · similarity 0.84)
- Learning from Synchronization: Self-Supervised Uncalibrated Multi-View Person Association in Challenging Scenes (CVPR 2025 · similarity 0.84)
- Motions as Queries: One-Stage Multi-Person Holistic Human Motion Capture (CVPR 2025 · similarity 0.84)
- Capturing Humans in Motion: Temporal-Attentive 3D Human Pose and Shape Estimation From Monocular Video (CVPR 2022 · similarity 0.83)
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.
