Research map / Detection and segmentation / CNN features
PossLoss: A Reliable and Sensitive Facial Landmark Detection Loss Function
Qikui Zhu
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 distribution-modulating term (2) to cross-entropy (ablation) | PossLoss · WFLW | NME: 4.27 -> 4.07 | Improved |
| add distribution-modulating term (2) to cross-entropy (ablation) | PossLoss · WFLW | FR10%: 2.88 -> 2.12 | Improved |
| add distribution-modulating term (2) to cross-entropy (ablation) | PossLoss · WFLW | AUC10%: 0.587 -> 0.600 | Improved |
| use PossLoss instead of Focal loss | Hourglass · WFLW | FR10%: 2.52 -> 2.12 | Improved |
| use Focal loss instead of cross-entropy | Hourglass · WFLW | NME: 4.27 -> 4.20 | Improved |
| use PossLoss instead of STAR loss | Hourglass · WFLW | NME: 4.02 -> 4.07 | Hurt |
| use PossLoss instead of STAR loss | Hourglass · WFLW | FR10%: 2.32 -> 2.12 | Improved |
| use PossLoss instead of STAR loss | Hourglass · 300W Challenging | NME: 4.32 -> 4.21 | Improved |
| use PossLoss instead of STAR loss | Hourglass · COFW | NME: 4.62 -> 4.60 | Improved |
| raise temperature d from 0.1 to 1.0 (ablation) | PossLoss · WFLW | NME: 4.07 -> 4.29 | Hurt |
| lower temperature d from 0.1 to 0.01 (ablation) | PossLoss · WFLW | NME: 4.07 -> 4.15 | Hurt |
| set weighting r=2 instead of r=1 (ablation) | PossLoss · WFLW | NME: 4.14 -> 4.07 | Improved |
Closest papers by meaning
- POPoS: Improving Efficient and Robust Facial Landmark Detection with Parallel Optimal Position Search (AAAI 2025 · similarity 0.90)
- Pixel-in-Pixel Net: Towards Efficient Facial Landmark Detection in the Wild (IJCV 2021 · similarity 0.88)
- Facial Landmark Detection: A Literature Survey (IJCV 2018 · similarity 0.87)
- Learning Robust Facial Landmark Detection via Hierarchical Structured Ensemble (ICCV 2019 · similarity 0.87)
- Contour-Hugging Heatmaps for Landmark Detection (CVPR 2022 · similarity 0.87)
- T-FAKE: Synthesizing Thermal Images for Facial Landmarking (CVPR 2025 · 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.
