Research map / Vision-language and multimodal / Cross-modal learning
Towards Unified Representation of Invariant-Specific Features in Missing Modality Face Anti-Spoofing
Guanghao Zheng, Yuchen Liu, Wenrui Dai, Chenglin Li, Junni Zou, Hongkai Xiong
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 modality-disentangle adapters (MDA) to MultiViT (ablation) | MMA-FAS · WMCA mean ACER | ACER (%): 12.39 -> 4.28 | Improved |
| add contrastive loss with batch/sample masking (ablation) | MMA-FAS · WMCA mean ACER | ACER (%): 4.28 -> 3.59 | Improved |
| add LBP-guided weighting to the contrastive loss (ablation) | MMA-FAS · WMCA mean ACER | ACER (%): 3.59 -> 2.72 | Improved |
| add adaptively modal combination sampling (AMS) (ablation) | MMA-FAS · WMCA mean ACER | ACER (%): 2.72 -> 2.26 | Improved |
| use a larger frequency mask kernel f (2 to 5) (ablation) | MDA · WMCA mean ACER | ACER (%): 2.26 -> 8.27 | Hurt |
| replace LBP with other feature descriptors (HoG, DoG, SIFT) in the contrastive loss (ablation) | MMA-FAS · WMCA mean ACER | ACER (%): 2.26 -> 3.99 (HoG); 3.18 (DoG); 2.82 (SIFT) | Hurt |
| vary adapter dimension K away from 64 (ablation) | MDA · WMCA mean ACER | ACER (%): 2.72 (K=64) vs 3.03 (16), 3.75 (32), 4.45 (78), 3.18 (128) | Mixed |
| apply adapters only to the FFN rather than also MHSA (ablation) | MMA-FAS · WMCA | ACER (%) | Improved |
| AMS sampling balances all weak modal combinations versus MMANet auxiliary classifier (ablation) | MMA-FAS · WMCA weak combinations during training | ACER | Improved |
| MMA-FAS compared with FM-ViT and MA-ViT on WMCA seen mean ACER | MMA-FAS · WMCA seen protocol | ACER (%): 2.54 -> 2.26 (FM-ViT) | Improved |
Closest papers by meaning
- Rethinking Vision Transformer and Masked Autoencoder in Multimodal Face Anti-Spoofing (IJCV 2024 · similarity 0.90)
- mmFAS: Multimodal Face Anti-Spoofing Using Multi-Level Alignment and Switch-Attention Fusion (AAAI 2025 · similarity 0.88)
- Suppress and Rebalance: Towards Generalized Multi-Modal Face Anti-Spoofing (CVPR 2024 · similarity 0.87)
- FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models (AAAI 2026 · similarity 0.86)
- Interpretable Face Anti-Spoofing: Enhancing Generalization with Multimodal Large Language Models (AAAI 2025 · similarity 0.86)
- From Intuition to Investigation: A Tool-Augmented Reasoning MLLM Framework for Generalizable Face Anti-Spoofing (CVPR 2026 · similarity 0.85)
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.
