Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
LUVLi Face Alignment: Estimating Landmarks' Location, Uncertainty, and Visibility Likelihood
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 |
|---|---|---|---|
| Use LUVLi rather than the strongest listed baseline | 300-W Split 2 with 300W-LP-2D pretraining · 300-W Test, Split 2 | NMEbox (%): 2.21 -> 2.10 | Helped |
| Use LUVLi rather than the strongest listed baseline | AFLW-19 face alignment · AFLW-19 Full | NMEdiag (%): 1.57 -> 1.39 | Helped |
| Use LUVLi rather than the strongest listed baseline | WFLW face alignment · WFLW-All | NMEocular (%): 4.36 -> 4.37 | No clear effect |
| Use LUVLi rather than the strongest listed baseline | MERL-RAV face alignment · MERL-RAV All | NMEbox (%): 1.99 -> 1.61 | Helped |
| apply loss to all hourglasses instead of only the last (ablation) | LUVLi · 300-W Test, Split 2 | NMEbox (%): 2.32 -> 2.10 | Helped |
| use LUVLi likelihood loss instead of MSE (ablation) | LUVLi · 300-W Test, Split 2 | NMEbox (%): 2.25 -> 2.10 | Helped |
| use Laplacian rather than Gaussian location likelihood (ablation) | LUVLi with visibility loss · 300-W Test, Split 2 | NMEbox (%): 2.13 -> 2.10 | Helped |
| include visibility loss (ablation) | LUVLi with Laplacian likelihood · 300-W Test, Split 2 | NMEbox (%): 2.10 -> 2.10 | No clear effect |
| initialize from 300W-LP-2D pretrained weights rather than scratch (ablation) | LUVLi · 300-W Test, Split 2 | NMEbox (%): 2.32 -> 2.10 | Helped |
| use heatmap spatial mean rather than direct regression (ablation) | LUVLi mean estimator · 300-W Test, Split 2 | NMEbox (%): 4.32 -> 2.10 | Helped |
| use ReLU rather than softmax heatmap weighting (ablation) | LUVLi mean estimator · 300-W Test, Split 2 | NMEbox (%): 2.37 -> 2.10 | Helped |
| use ReLU rather than temperature-scaled softmax heatmap weighting (ablation) | LUVLi mean estimator · 300-W Test, Split 2 | NMEbox (%): 2.10 -> 2.10 | No clear effect |
Closest papers by meaning
- Look at Boundary: A Boundary-Aware Face Alignment Algorithm (CVPR 2018 · similarity 0.89)
- Brute-Force Facial Landmark Analysis With a 140,000-Way Classifier (AAAI 2018 · similarity 0.89)
- Semantic Alignment: Finding Semantically Consistent Ground-Truth for Facial Landmark Detection (CVPR 2019 · similarity 0.89)
- AnchorFace: An Anchor-based Facial Landmark Detector Across Large Poses (AAAI 2021 · similarity 0.88)
- FaceLift: Semi-supervised 3D Facial Landmark Localization (CVPR 2024 · similarity 0.87)
- Generalizable Face Landmarking Guided by Conditional Face Warping (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.