Research map / Label-efficient and robust learning / Self-supervised and contrastive
Rethinking Vision Transformer and Masked Autoencoder in Multimodal Face Anti-Spoofing
Zitong Yu, Rizhao Cai, Yawen Cui, Xin Liu, Yongjian Hu, Alex Chichung Kot
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 AMA adapter finetuning instead of direct finetuning | ViT-Base · WMCA unseen, RGB+IR+Depth | ACER mean (%): 6.46 -> 5.56 | Improved |
| add AMA adapter finetuning | ViT-Base · MmFA RGB+Depth | ACER (%): 4.60 -> 1.84 | Improved |
| add ConvAdapter finetuning | ViT-Base · MmFA RGB | ACER (%): 17.37 -> 10.02 | Improved |
| add M2A2E pre-training to AMA | ViT+AMA · WMCA unseen, RGB+IR+Depth | ACER mean (%): 5.56 -> 3.86 | Improved |
| add M2A2E pre-training to ConvA | ViT+ConvA · WMCA unseen, RGB | ACER mean (%): 9.42 -> 11.5 | Hurt |
| add M2A2E pre-training to AMA | ViT+AMA · WMCA unseen, RGB+Depth | ACER mean (%): 5.78 -> 5.82 | No clear effect |
| add M2A2E pre-training to AMA | ViT+AMA · Cross-dataset WMCA->MmFA, RGB+IR+Depth | HTER (%): 13.99 -> 8.60 | Improved |
| add M2A2E pre-training to ConvA | ViT+ConvA · MmFA intra-test, RGB | ACER (%): 10.02 -> 6.40 | Improved |
| add M2A2E pre-training to ConvA | ViT+ConvA · Cross-dataset WMCA->MmFA, RGB | HTER (%): 21.31 -> 23.56 | Hurt |
| add adaptive modality weighting (AMA) over multimodal ConvAdapter (ablation) | multimodal ConvAdapter · WMCA seen, multimodal | ACER (%): decreases 0.52%, 0.31%, 1.07%, 0.07% | Improved |
| use multimodal ConvAdapter instead of independent ConvAdapter (ablation) | ConvAdapter · WMCA seen, multimodal | ACER (%): reduces more than 0.5% ACER | Improved |
| use huge multimodal ConvAdapter (D'xK output) (ablation) | multimodal ConvAdapter · WMCA seen, multimodal | ACER (%) | No clear effect |
Closest papers by meaning
- mmFAS: Multimodal Face Anti-Spoofing Using Multi-Level Alignment and Switch-Attention Fusion (AAAI 2025 · similarity 0.90)
- Interpretable Face Anti-Spoofing: Enhancing Generalization with Multimodal Large Language Models (AAAI 2025 · similarity 0.89)
- FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models (AAAI 2026 · similarity 0.88)
- PA-FAS: Towards Interpretable and Generalizable Multimodal Face Anti-Spoofing via Path-Augmented Reinforcement Learning (AAAI 2026 · similarity 0.88)
- FSFM: A Generalizable Face Security Foundation Model via Self-Supervised Facial Representation Learning (CVPR 2025 · similarity 0.87)
- SLIP: Spoof-Aware One-Class Face Anti-Spoofing with Language Image Pretraining (AAAI 2025 · 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.
