Research map / Architectures and efficiency / Attention and transformers
Sparse Local Patch Transformer for Robust Face Alignment and Landmarks Inherent Relation Learning
Check 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 coarse-to-fine localization with 3 stages instead of 1 (ablation) | SLPT · WFLW test | NME (%): 4.79% -> 4.14% | Helped |
| Add MSA and MCA attention blocks (ablation) | SLPT · WFLW test | NME (%): 4.48% -> 4.14% | Helped |
| Add MSA self-attention (ablation) | SLPT with MCA removed · WFLW test | NME (%): 4.48% -> 4.20% | Helped |
| Add MCA cross-attention (ablation) | SLPT with MSA removed · WFLW test | NME (%): 4.48% -> 4.17% | Helped |
| Add structure encoding (ablation) | SLPT · WFLW test | NME (%): 4.16% -> 4.14% | Helped |
| Use 5 × 5 patches instead of 7 × 7 patches (ablation) | SLPT · WFLW test | NME (%): 4.14% -> 4.17% | Hurt |
| Use 9 × 9 patches instead of 7 × 7 patches (ablation) | SLPT · WFLW test | NME (%): 4.14% -> 4.16% | Hurt |
| Increase inherent relation layers from 2 to 12 (ablation) | SLPT · WFLW test | NME (%): 4.19% -> 4.12% | Helped |
| Increase coarse-to-fine stages beyond 3 (ablation) | SLPT · WFLW test | NME (%): 4.14% -> 4.24% | No clear effect |
| Use 12 rather than 6 relation layers (ablation) | SLPT · WFLW test | NME (%): 4.14% -> 4.12% | Helped |
Closest papers by meaning
- FreeEnricher: Enriching Face Landmarks without Additional Cost (AAAI 2023 · similarity 0.87)
- Disentangling 3D Pose in a Dendritic CNN for Unconstrained 2D Face Alignment (CVPR 2018 · similarity 0.86)
- Self-Supervised Facial Representation Learning with Facial Region Awareness (CVPR 2024 · similarity 0.84)
- T-FAKE: Synthesizing Thermal Images for Facial Landmarking (CVPR 2025 · similarity 0.84)
- SCE-MAE: Selective Correspondence Enhancement with Masked Autoencoder for Self-Supervised Landmark Estimation (CVPR 2024 · similarity 0.84)
- LAFS: Landmark-based Facial Self-supervised Learning for Face Recognition (CVPR 2024 · similarity 0.84)
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.