Research map / Label-efficient and robust learning / Recognition and long-tail
SDD-FIQA: Unsupervised Face Image Quality Assessment With Similarity Distribution Distance
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 Wasserstein distance between positive- and negative-pair similarity distributions as quality labels | SDD-FIQA · LFW, Adience, and IJB-C | AOC: 0.7149 -> 0.8149; 0.5698 -> 0.5996; 0.7157 -> 0.7329 | Helped |
| Use Wasserstein distance between positive- and negative-pair similarity distributions as quality labels | SDD-FIQA · LFW, Adience, and IJB-C with deployed ResNet101-MS1M | AOC: 0.6790 -> 0.7979; 0.5736 -> 0.6114; 0.7109 -> 0.7301 | Helped |
| Use Wasserstein distance between positive- and negative-pair similarity distributions as quality labels | SDD-FIQA · LFW, Adience, and IJB-C with deployed ResNet101-CASIA | AOC: 0.6369 -> 0.6898; 0.4425 -> 0.4589; 0.5678 -> 0.5743 | Helped |
| Use SDD-FIQA quality scores to weight face samples in set aggregation | IJB-C set-based verification · IJB-C | TAR: 0.9473 -> 0.9617 | Helped |
| Use both positive- and negative-pair similarity distributions instead of only one (ablation) | SDD-FIQA pseudo-label generation · LFW | FNMR | Helped |
| Use positive-pair similarity information only, rather than negative-pair information only (ablation) | SDD-FIQA pseudo-label generation · LFW | FNMR | Helped |
| Increase the number of sampled positive and negative pairs m (ablation) | SDD-FIQA pseudo-label generation · LFW, Adience, and IJB-C | AOC | Helped |
| Increase the number of repeated label calculations K (ablation) | SDD-FIQA pseudo-label generation · LFW, Adience, and IJB-C | AOC | Helped |
Closest papers by meaning
- CLIB-FIQA: Face Image Quality Assessment with Confidence Calibration (CVPR 2024 · similarity 0.91)
- DSL-FIQA: Assessing Facial Image Quality via Dual-Set Degradation Learning and Landmark-Guided Transformer (CVPR 2024 · similarity 0.88)
- RepFace: Refining Closed-Set Noise with Progressive Label Correction for Face Recognition (AAAI 2025 · similarity 0.84)
- Assessing Uncertainty in Similarity Scoring: Performance & Fairness in Face Recognition (ICLR 2024 · similarity 0.84)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.84)
- UIFace: Unleashing Inherent Model Capabilities to Enhance Intra-Class Diversity in Synthetic Face Recognition (ICLR 2025 · similarity 0.83)
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.