Research map / Label-efficient and robust learning / Semi-supervised learning
Improving Landmark Localization With Semi-Supervised Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 ELT loss to sequential landmark training (ablation) | Seq-MT on HGR1 · HGR1 test, 5% labeled | normalized landmark error: 57.6 -> 43.7 | Helped |
| Add auxiliary gesture labels to landmark training with ELT (ablation) | Seq-MT on HGR1 · HGR1 test, 5% labeled | normalized landmark error: 43.7 -> 38.5 | Helped |
| Add ELT loss to sequential landmark training (ablation) | Seq-MT on Blocks · Blocks test, 5% labeled | pixel error: 3.95 -> 1.94 | Helped |
| Add auxiliary class labels to sequential landmark training with ELT (ablation) | Seq-MT on Blocks · Blocks test, 5% labeled | pixel error: 1.94 -> 1.70 | Helped |
| Add auxiliary camera and emotion labels (ablation) | Seq-MT on Multi-PIE · Multi-PIE test, 5% labeled | normalized landmark error: 6.69 -> 6.57 | Helped |
| Add ELT loss to RCN+ landmark training (ablation) | RCN+ on 300W · 300W Fullset test, 100% labeled | normalized landmark error: 5.26 -> 5.10 | Helped |
| Add ELT loss and auxiliary pose regression (ablation) | RCN+ on AFLW · AFLW test, 5% labeled | normalized landmark error: 2.17 -> 2.03 | Helped |
| Replace softmax landmark decoding with soft-argmax (ablation) | Heatmap-MT (L+A) · Multi-PIE test, 5% labeled | normalized landmark error: 11.03 -> 8.00 | Helped |
| Replace softmax landmark decoding with soft-argmax (ablation) | Heatmap-MT (L+A) · HGR1 test, 5% labeled | normalized landmark error: 64.8 -> 56.88 | Helped |
| Use sequential landmark bottleneck rather than common multitask branches | Landmark localization on Blocks · Blocks test, 5% labeled | pixel error: 11.56 -> 3.95 | Helped |
| Use sequential landmark bottleneck rather than heatmap multitasking | Landmark localization on Blocks · Blocks test, 5% labeled | pixel error: 6.59 -> 3.95 | Helped |
Closest papers by meaning
- FaceLift: Semi-supervised 3D Facial Landmark Localization (CVPR 2024 · similarity 0.89)
- AnchorFace: An Anchor-based Facial Landmark Detector Across Large Poses (AAAI 2021 · similarity 0.88)
- Few-Shot Geometry-Aware Keypoint Localization (CVPR 2023 · similarity 0.88)
- Learning to Localize Little Landmarks (CVPR 2016 · similarity 0.88)
- Brute-Force Facial Landmark Analysis With a 140,000-Way Classifier (AAAI 2018 · similarity 0.87)
- POPoS: Improving Efficient and Robust Facial Landmark Detection with Parallel Optimal Position Search (AAAI 2025 · similarity 0.85)
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.