Research map / 3D vision / Depth and stereo
Supervision by Registration and Triangulation for Landmark Detection
Xuanyi Dong, Yi Yang, Shih-En Wei, Xinshuo Weng, Yaser Ajmal Sheikh, Shoou-I Yu
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 (loss weight 1.0) | regression-based detector · Mugsy-V1 test | NME: 4.63 -> 4.44 | Improved |
| add SBT (loss weight 0.5) | regression-based detector · Mugsy-V1 test | P-error: 5.14 -> 3.80 | Improved |
| add SBR+SBT jointly (0.5/0.5) | regression-based detector · Mugsy-V1 test | NME: 4.63 -> 4.36 | Improved |
| raise SBT weight to 2.0 (ablation) | regression-based detector · Mugsy-V1 test | NME: 4.39 -> 4.60 | Hurt |
| use L1 loss on coordinates instead of L2 for SBT (ablation) | regression-based detector · Mugsy-V1 test | P-error: 4.70 -> 3.38 | Improved |
| use L1 loss on coordinates for SBR (ablation) | heatmap-based detector · Mugsy-V1 test | NME: 4.04 -> 7.67 | Hurt |
| use L2 on heatmaps instead of coordinates for SBT (ablation) | heatmap-based detector · Mugsy-V1 test | NME: 3.98 -> 3.80 | Improved |
| add SRT to heatmap detector with 68 landmarks | heatmap-based detector · 300-W Full (68 landmarks) | NME: 3.34 -> 3.39 | Hurt |
| add SRT to regression detector with 49 landmarks | regression-based detector · 300-W Full (49 landmarks) | NME: 3.55 -> 3.31 | Improved |
| add SRT to heatmap detector | heatmap-based detector · 300-W Full (49 landmarks) | P-error: 9.94 -> 8.87 | Improved |
| use VoxCeleb2 instead of 300-VW as unlabeled data (ablation) | regression-based detector · 300-W Full | NME: 3.39 -> 3.36 | Hurt |
| add SRT to heatmap detector | heatmap-based detector · AFLW-Front | NME: 1.71 -> 1.64 | Improved |
Closest papers by meaning
- Supervision-by-Registration: An Unsupervised Approach to Improve the Precision of Facial Landmark Detectors (CVPR 2018 · similarity 0.90)
- Unsupervised Learning of Object Landmarks via Self-Training Correspondence (NeurIPS 2020 · similarity 0.87)
- Mining and Transferring Feature-Geometry Coherence for Unsupervised Point Cloud Registration (NeurIPS 2024 · similarity 0.85)
- Scalable 3D Registration via Truncated Entry-wise Absolute Residuals (CVPR 2024 · similarity 0.84)
- Pose-Guided Self-Training with Two-Stage Clustering for Unsupervised Landmark Discovery (CVPR 2024 · similarity 0.84)
- Learning from Synchronization: Self-Supervised Uncalibrated Multi-View Person Association in Challenging Scenes (CVPR 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.
