Research map / Label-efficient and robust learning / Domain adaptation and generalization
PriSM: Parsing and Style-Mixed Consistency for Unsupervised Domain Adaptation in Facial Landmark Detection
Chieh-Yu Yang, Hou-Ning Hu, Sykai Chen, Yu-Lun Liu, Yen-Yu Lin
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 parsing consistency loss (CE) to source-only detector (ablation) | PriSM · CariFace UDA | NME: 11.14 -> 7.41 | Improved |
| add LPIPS reconstruction loss next to parsing CE loss (ablation) | PriSM · CariFace UDA | NME: 7.41 -> 7.35 | Improved |
| add parsing CE loss to LPIPS reconstruction loss (ablation) | PriSM · ArtiFace UDA | NME: 3.70 -> 3.61 | Improved |
| use LPIPS instead of DINO for reconstruction loss (ablation) | PriSM · ArtiFace UDA | NME: 3.93 -> 3.70 | Improved |
| use LPIPS instead of Sobel gradient loss for reconstruction (ablation) | PriSM · CariFace UDA | NME: 7.71 -> 8.07 | Hurt |
| use latent optimization instead of encoder-based GAN inversion (ablation) | PriSM · ArtiFace UDA | NME: 3.73 -> 3.61 | Improved |
| use latent optimization instead of encoder-based GAN inversion (ablation) | PriSM · CariFace UDA | NME: 7.37 -> 7.35 | No clear effect |
| PriSM vs CFW on 300W->CariFace | PriSM · CariFace UDA | NME: 7.70 -> 7.35 | Improved |
| PriSM vs CFW on 300W->ArtiFace | PriSM · ArtiFace UDA | NME: 3.93 -> 3.61 | Improved |
| PriSM vs CFW, unseen CariFace after ArtiFace adaptation | PriSM · CariFace unseen | NME: 10.93 -> 10.92 | No clear effect |
| adapt landmarks with PriSM vs unadapted source landmarks, parsing quality | PriSM · CariFace parsing | mIoU: 36.83 -> 38.37 | Improved |
Closest papers by meaning
- Aggregation via Separation: Boosting Facial Landmark Detector With Semi-Supervised Style Translation (ICCV 2019 · similarity 0.90)
- Semantic Alignment: Finding Semantically Consistent Ground-Truth for Facial Landmark Detection (CVPR 2019 · similarity 0.88)
- Style Aggregated Network for Facial Landmark Detection (CVPR 2018 · similarity 0.88)
- Generalizable Face Landmarking Guided by Conditional Face Warping (CVPR 2024 · similarity 0.88)
- Self-Reinforced Cascaded Regression for Face Alignment (AAAI 2018 · similarity 0.88)
- LAFS: Landmark-based Facial Self-supervised Learning for Face Recognition (CVPR 2024 · similarity 0.87)
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.
