Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain Generalization via Ensemble Stacking for Face Presentation Attack Detection
Usman Muhammad, Jorma T. Laaksonen, Djamila Romaissa Beddiar, Mourad Oussalah
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 synthetic motion-blur data and an LSTM to the CNN (ablation) | CNN · O&C&I to M | HTER: 23.70 -> 8.65 | Improved |
| add synthetic data and an LSTM to the CNN (ablation) | CNN · I&C&M to O | HTER: 37.55 -> 10.60 | Improved |
| add synthetic data and an LSTM to the CNN (ablation) | CNN · O&C&M to I | HTER: 39.44 -> 38.37 | No clear effect |
| add GRU meta-model stacking over three base models (ablation) | best single base model · O&C&I to M | HTER: 3.95 -> 4.81 | Mixed |
| add GRU meta-model stacking over three base models (ablation) | base model 3 (GRU) · O&C&I to M | HTER: 14.44 -> 4.81 | Improved |
| add GRU meta-model stacking over three base models (ablation) | base model 2 (BiLSTM) · I&C&M to O | HTER: 25.41 -> 6.70 | Improved |
| add GRU meta-model stacking over three base models (ablation) | base model 1 (LSTM) · I&C&M to O | HTER: 5.35 -> 6.70 | Hurt |
| use the stacked meta-model against the best prior method | Meta-Model vs Regression Network · O&C&M to I (Idiap) | HTER: 8.62 -> 38.11 | Hurt |
| use the stacked meta-model against ADL | Meta-Model vs ADL · O&C&I to M | HTER: 5.00 -> 4.81 | Improved |
| use the stacked meta-model with only two source domains | Meta-Model vs EPCR-unlabeled · O&I to M | HTER: 10.4 -> 8.1 | Improved |
| use the proposed FAST+FREAK global motion estimation instead of TSS | TSS method · CASIA-FASD | execution time (s): 1140 -> 1023 | Improved |
| use the proposed global motion estimation instead of optical flow | optical flow · REPLAY-ATTACK | execution time (s): 1082 -> 641 | Improved |
Closest papers by meaning
- Multi-Adversarial Discriminative Deep Domain Generalization for Face Presentation Attack Detection (CVPR 2019 · similarity 0.91)
- Test-Time Domain Generalization for Face Anti-Spoofing (CVPR 2024 · similarity 0.90)
- Multi-Domain Incremental Learning for Face Presentation Attack Detection (AAAI 2024 · similarity 0.88)
- Gradient Alignment for Cross-Domain Face Anti-Spoofing (CVPR 2024 · similarity 0.87)
- Rethinking Generalizable Face Anti-spoofing via Hierarchical Prototype-guided Distribution Refinement in Hyperbolic Space (CVPR 2024 · similarity 0.87)
- FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models (AAAI 2026 · 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.
