Research map / Vision-language and multimodal / Multimodal LLMs
TF-FAS: Twofold-Element Fine-Grained Semantic Guidance for Generalizable Face Anti-Spoofing
Xudong Wang, Ke-Yue Zhang, Taiping Yao, Qianyu Zhou, Shouhong Ding, Pingyang Dai, Rongrong Ji
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 FCEM (fine-grained categorical element module) (ablation) | dual-stream CLIP baseline · C->I, C->M, C->O average | HTER: 17.22 -> 9.39 | Improved |
| add CEDM (content element decoupling module) (ablation) | baseline + FCEM · C->I, C->M, C->O average | HTER: 9.39 -> 6.14 | Improved |
| use fine-grained instead of coarse category prompts (ablation) | FCEM · C->I, C->M, C->O average | HTER: 17.22 -> 13.87 | Improved |
| learn fine-grained prompts with CoOp instead of manual prompts (ablation) | FCEM · C->I, C->M, C->O average | HTER: 13.87 -> 14.56 | Hurt |
| expand category prompts via GPT-4 augmentation (ablation) | FCEM · C->I, C->M, C->O average | HTER: 13.87 -> 11.55 | Improved |
| add filtering of augmented samples (ablation) | FCEM · C->I, C->M, C->O average | HTER: 11.55 -> 10.82 | Improved |
| add adaptive weight learning for prompts (ablation) | FCEM · C->I, C->M, C->O average | HTER: 10.82 -> 9.39 | Improved |
| use orthogonality disentangling loss instead of CLUB, MINE or contrastive (ablation) | CEDM · C->I, C->M, C->O average | HTER: 7.58 (MINE) -> 6.14 | Improved |
| disentangle both environment and face semantics instead of one (ablation) | CEDM · C->I, C->M, C->O average | HTER: 6.50 (face only) -> 6.14 (all) | Mixed |
| add CelebA-Spoof as extra source data | TF-FAS · Protocol 1 average | HTER: 2.19 -> 1.27 | Improved |
Closest papers by meaning
- Interpretable Face Anti-Spoofing: Enhancing Generalization with Multimodal Large Language Models (AAAI 2025 · similarity 0.90)
- From Intuition to Investigation: A Tool-Augmented Reasoning MLLM Framework for Generalizable Face Anti-Spoofing (CVPR 2026 · similarity 0.90)
- CFPL-FAS: Class Free Prompt Learning for Generalizable Face Anti-spoofing (CVPR 2024 · similarity 0.90)
- Test-Time Domain Generalization for Face Anti-Spoofing (CVPR 2024 · similarity 0.90)
- FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models (AAAI 2026 · similarity 0.89)
- Gradient Alignment for Cross-Domain Face Anti-Spoofing (CVPR 2024 · similarity 0.89)
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.
