Research map / Video / Video segmentation and tracking
Supervision-by-Registration: An Unsupervised Approach to Improve the Precision of Facial Landmark Detectors
Xuanyi Dong, Shoou-I Yu, Xinshuo Weng, Shih-En Wei, Yi Yang, Yaser Sheikh
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 SBR using unlabeled video | CPM · 300-W full | NME: 4.36 -> 4.10 | Improved |
| Add SBR using unlabeled video | CPM · AFLW | NME: 2.33 -> 2.14 | Improved |
| Add SBR using unlabeled video | CPM · 300-VW category C | AUC@0.08: 57.25 -> 58.22 | Improved |
| Add PAM adaptation on the video | CPM + SBR · 300-VW category C | AUC@0.08: 58.22 -> 59.39 | Improved |
| Adapt the detector to the testing video with PAM | CPM · YouTube Celebrities | NME: 5.26 -> 4.74 | Improved |
| Adapt the detector to the testing video with PAM | Reg · YouTube Celebrities | NME: 10.21 -> 9.31 | Improved |
| Use 5-frame rather than 3-frame LK tracking (ablation) | CPM + PAM · YouTube Celebrities | NME: 4.74 -> 5.01 | Hurt |
| Use conv-1 features rather than RGB images for LK tracking (ablation) | CPM + PAM · YouTube Celebrities | NME: 4.74 -> 5.13 | Hurt |
| Use self-training without registration-loss feedback (ablation) | CPM · YouTube Celebrities | NME: 4.74 -> 5.45 | Hurt |
Closest papers by meaning
- Supervision by Registration and Triangulation for Landmark Detection (TPAMI 2020 · similarity 0.90)
- Recurrence Without Recurrence: Stable Video Landmark Detection With Deep Equilibrium Models (CVPR 2023 · similarity 0.87)
- Synergy Between Face Alignment and Tracking via Discriminative Global Consensus Optimization (ICCV 2017 · similarity 0.86)
- Teacher Supervises Students How to Learn From Partially Labeled Images for Facial Landmark Detection (ICCV 2019 · similarity 0.86)
- Tracktention: Leveraging Point Tracking to Attend Videos Faster and Better (CVPR 2025 · similarity 0.83)
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · 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.
