Research map / Image generation / Editing and style transfer
Style Aggregated Network for Facial Landmark Detection
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 |
|---|---|---|---|
| Use SAN with official detector bounding boxes instead of Two-Stage_OD | 300-W landmark detection · 300-W Common | NME: 4.36 -> 3.41 | Helped |
| Use SAN with official detector bounding boxes instead of Two-Stage_GT | 300-W landmark detection · 300-W Full Set | NME: 4.96 -> 4.24 | Helped |
| Use SAN instead of Two-Stage | AFLW landmark detection · AFLW-Full | NME: 2.17 -> 1.91 | Helped |
| Use SAN instead of CCL | AFLW landmark detection · AFLW-Front | NME: 2.17 -> 1.85 | Helped |
| Add GAN-generated style-aggregated face input to the detector (ablation) | SAN without GAN with augmented training data · AFLW-Full | AUC@0.08: 78.99 -> 79.82 | Helped |
| Add a GAN-generated style-aggregated input alongside the original image (ablation) | SAN without GAN · 300W-Style Common | NME: 4.06 -> 3.98 | Helped |
| Add a GAN-generated style-aggregated input alongside the original image (ablation) | SAN without GAN · 300W-Style Challenging | NME: 6.88 -> 6.60 | Helped |
| Add a GAN-generated style-aggregated input alongside the original image (ablation) | SAN without GAN · 300W-Style Full | NME: 4.06 -> 3.98 | Helped |
| Replace the style-aggregated stream with light-style images (ablation) | SAN · 300-W Challenging, Gray test | NME: 6.91 -> 7.30 | Hurt |
| Use ground-truth face bounding boxes instead of official detector boxes | SAN · 300-W Challenging | NME: 7.55 -> 6.60 | Helped |
| Use style aggregation when training and testing styles differ | SAN · 300W-Style Full | NME: On average, SAN obtains 7% relative improvement | Helped |
| Increase the number of style clusters and generation models beyond k=3 (ablation) | SAN style-aggregated face generation · AFLW | NME | No clear effect |
Closest papers by meaning
- Brute-Force Facial Landmark Analysis With a 140,000-Way Classifier (AAAI 2018 · similarity 0.88)
- Semantic Alignment: Finding Semantically Consistent Ground-Truth for Facial Landmark Detection (CVPR 2019 · similarity 0.87)
- Robust Facial Landmark Detection via a Fully-Convolutional Local-Global Context Network (CVPR 2018 · similarity 0.87)
- Unsupervised Learning of Object Landmarks through Conditional Image Generation (NeurIPS 2018 · similarity 0.87)
- Generalizable Face Landmarking Guided by Conditional Face Warping (CVPR 2024 · similarity 0.85)
- 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.