Research map / 3D vision / Human pose estimation
Self-supervised Multi-view Person Association and Its Applications
Minh Phuoc Vo, Ersin Yumer, Kalyan Sunkavalli, Sunil Hadap, Yaser Ajmal Sheikh, Srinivasa G. Narasimhan
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 keypoint and PAF heatmaps to RGB input (ablation) | Big RGB · ViPER (ReID) | top-1 accuracy: 42.4 -> 52.5 | Improved |
| add keypoint and PAF heatmaps to RGB input (ablation) | Big RGB · CUHK03 (ReID) | top-1 accuracy: 91.1 -> 93.7 | Improved |
| add keypoint and PAF heatmaps to RGB input (ablation) | Big RGB · MARS (ReID) | top-1 accuracy: 76.9 -> 79.8 | Improved |
| use both KP and PAF instead of KP alone (ablation) | Big RGB+KP · PRID (ReID) | top-1 accuracy: 60.0 -> 62.0 | Improved |
| adapt descriptor with multitask learning | pretrained descriptor · Chasing [C] frequent actors | 1-NN accuracy: 68% -> 94% | Improved |
| adapt descriptor with multitask learning | pretrained descriptor · Tagging [T] frequent actors | 1-NN accuracy: 75% -> 90% | Improved |
| finetune only on domain triplets (no multitask) (ablation) | pretrained descriptor · Chasing [C] camera holders | 1-NN accuracy: 85% -> 56% | Hurt |
| add labeled ReID classification loss to finetuning (MTL) (ablation) | triplet-only finetuning · Chasing [C] camera holders | 1-NN accuracy: 56% -> 95% | Improved |
| use multitask learning over pretrained model | pretrained descriptor · Chasing [C], Tagging [T], Halloween [H] | association accuracy: 17% [C], 9% [T], 9% [H] | Improved |
| use continuous clustering instead of k-means (ablation) | Kmeans+MTL · Tagging [T] | clustering accuracy: 88.2 -> 95.1 | Improved |
| use MTL descriptor instead of pretrained with same clustering (ablation) | [cite]+P · Halloween [H] | clustering accuracy: 79.5 -> 88.1 | Improved |
| add multi-view tracked 3D skeleton optimization over per-frame triangulation | Baseline · Chasing [C] | length deviation (cm): 8.0 -> 1.5 | Improved |
Closest papers by meaning
- Learning from Synchronization: Self-Supervised Uncalibrated Multi-View Person Association in Challenging Scenes (CVPR 2025 · similarity 0.90)
- Unveiling the Power of Self-Supervision for Multi-View Multi-Human Association and Tracking (TPAMI 2024 · similarity 0.90)
- ArtTrack: Articulated Multi-Person Tracking in the Wild (CVPR 2017 · similarity 0.90)
- CoMotion: Concurrent Multi-person 3D Motion (ICLR 2025 · similarity 0.88)
- Motions as Queries: One-Stage Multi-Person Holistic Human Motion Capture (CVPR 2025 · similarity 0.88)
- Self-Supervised Multi-Object Tracking with Path Consistency (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.
