Research map / Image generation / Editing and style transfer
MR-FIQA: Face Image Quality Assessment with Multi-Reference Representations from Synthetic Data Generation
Fu-Zhao Ou, Chongyi Li, Shiqi Wang, Sam Kwong
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 occlusion control via text prompts (ablation) | SynFIQA generation · avg of CPLFW, Adience, TinyFace | pAUC: 0.792 -> 0.781 | Improved |
| add blur/downsampling post-processing (ablation) | SynFIQA generation with OC · avg of CPLFW, Adience, TinyFace | pAUC: 0.781 -> 0.738 | Improved |
| add MR-FIQA quality labels in place of CR-FIQA training labels (ablation) | SynFIQA · 8 test sets avg | pAUC: 0.800 -> 0.730 | Improved |
| add MR-FIQA quality labels in place of CR-FIQA training labels (ablation) | SynFIQA · 8 test sets avg | AUC: 0.554 -> 0.520 | Improved |
| use MR-FIQA instead of CR-FIQA labels (ablation) | MobileFaceNet FIQA on SynFIQA · avg of 4 test sets | avg: 0.596 -> 0.572 | Improved |
| use FaceQnet instead of SDD-FIQA labels (ablation) | MobileFaceNet FIQA on SynFIQA · avg of 4 test sets | avg: 0.652 -> 0.643 | Improved |
| add spatial-domain score sd to recognition-domain fr (ablation) | MR-FIQA · avg of CPLFW, Adience, TinyFace | pAUC: 0.765 -> 0.723 | Improved |
| add vision-language score vl to recognition-domain fr (ablation) | MR-FIQA · avg of CPLFW, Adience, TinyFace | pAUC: 0.765 -> 0.731 | Improved |
| add vision-language score vl to fr+sd (ablation) | MR-FIQA · avg of CPLFW, Adience, TinyFace | pAUC: 0.723 -> 0.683 | Improved |
| use fr alone vs sd alone vs vl alone (ablation) | MR-FIQA · avg of CPLFW, Adience, TinyFace | pAUC: 0.765 vs 0.816 vs 0.858 | Mixed |
| train on SynFIQA instead of real CASIA-WebFace (SynFIQA++ vs CASIA) | CR-FIQA IR50 · 8 test sets avg | pAUC: 0.701 -> 0.730 | Hurt |
| combine SynFIQA with CASIA-WebFace (ablation) | CR-FIQA IR50 · LFW, XQLFW, AgeDB, CP-LFW, CFP-FP | pAUC | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DCFace — DCFace: Synthetic Face Generation With Dual Condition Diffusion Model (CVPR 2023) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- CLIB-FIQA: Face Image Quality Assessment with Confidence Calibration (CVPR 2024 · similarity 0.92)
- DSL-FIQA: Assessing Facial Image Quality via Dual-Set Degradation Learning and Landmark-Guided Transformer (CVPR 2024 · similarity 0.90)
- UIFace: Unleashing Inherent Model Capabilities to Enhance Intra-Class Diversity in Synthetic Face Recognition (ICLR 2025 · similarity 0.85)
- Toward Generalized Image Quality Assessment: Relaxing the Perfect Reference Quality Assumption (CVPR 2025 · similarity 0.85)
- Towards Syn-to-Real IQA: A Novel Perspective on Reshaping Synthetic Data Distributions (NeurIPS 2025 · similarity 0.84)
- $\text{ID}^3$: Identity-Preserving-yet-Diversified Diffusion Models for Synthetic Face Recognition (NeurIPS 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.
