EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Dynamic Cross-Modal Modeling with an Ultra-Lightweight Architecture for Face Anti-Spoofing

Nana Li iD, Jiayu Wang, Zuhe Li iD, Zhipeng Weng, Yihong Wang, Yushan Pan iD

Sensors · 27.04.2026 · 26 (9) · 2699

DOI: 10.3390/s26092699 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 27.04.2026. Verfasst von Nana Li und 5 weiteren Autor:innen. Beteiligte Einrichtung: School of Computer Science and Artificial Intelligence, Zhengzhou University of Light Industry, Zhengzhou 450002, China. Die Arbeit nennt 48 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Henan Science and Technology Department, Zhengzhou University of Light Industry, Jiangsu University of Science and Technology.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The effectiveness of multimodal face anti-spoofing largely depends on the modeling of cross-modal relationships. However, most existing approaches rely on static fusion or implicitly learned feature aggregation, which assumes fixed modality importance, limiting its ability to capture reliability variations across different attack patterns. Under strict computational constraints, achieving effective dynamic cross-modal modeling remains a significant challenge. To address this issue, we propose an ultra-lightweight dynamic cross-modal framework for face anti-spoofing, with ultra-low parameters, FLOPs, latency, memory and high FPS for real-time edge inference. A compact feature extractor is constructed by enhancing ShuffleNetV2 with the Ghost-Generated Shuffle BlockA (GGS-BlockA), which significantly reduces redundant computation while maintaining high discriminative capability. On this basis, a Lightweight Cross-Modal Attention (LCMA) module performs sample-wise dynamic modality reweighting to capture reliability variations among RGB, Depth, and IR modalities. Furthermore, a Lightweight Cross-Modal Fusion (LCMF) module utilizes depth cues as stable guidance to improve cross-modal feature alignment and complementary representation. Experiments on the CASIA-SURF benchmark demonstrate that the proposed method achieves an Average Classification Error Rate (ACER) of 0.064% with only 0.14M parameters and 0.0065G FLOPs. At the strict threshold of TPR@FPR=10−4, a detection rate of 99.86% is obtained, demonstrating strong robustness and generalization capability under extremely low computational cost.

Schlagwörter

Stichwörter der Autor:innen

dynamic cross-modal modelingface anti-spoofinglightweight architecturemultimodal fusion

Autor:innen und Institutionen

  • School of Computer Science and Artificial Intelligence, Zhengzhou University of Light Industry, Zhengzhou 450002, China.
  • Department of Computing, Xi'an Jiaotong-Liverpool University, Suzhou 215123, China.

ORCID-Profile: Nana Li, Zuhe Li, Yushan Pan

Förderung

  • Henan Science and Technology Department
  • Zhengzhou University of Light Industry
  • Jiangsu University of Science and Technology
  • Henan Provincial Science and Technology Research Project
  • Jiangsu Science and Technology Program

Zitieren

APA

Li, N., Wang, J., Li, Z., Weng, Z., Wang, Y., & Pan, Y. (2026). Dynamic Cross-Modal Modeling with an Ultra-Lightweight Architecture for Face Anti-Spoofing. Sensors, 26(9), 2699. https://doi.org/10.3390/s26092699

Vancouver

Li N, Wang J, Li Z, Weng Z, Wang Y, Pan Y. Dynamic Cross-Modal Modeling with an Ultra-Lightweight Architecture for Face Anti-Spoofing. Sensors. 2026;26(9):2699. doi:10.3390/s26092699

BibTeX
@article{Li2026,
  author = {Li, Nana and Wang, Jiayu and Li, Zuhe and Weng, Zhipeng and Wang, Yihong and Pan, Yushan},
  title = {Dynamic Cross-Modal Modeling with an Ultra-Lightweight Architecture for Face Anti-Spoofing},
  journal = {Sensors},
  year = {2026},
  volume = {26},
  number = {9},
  pages = {2699},
  doi = {10.3390/s26092699}
}
Kostenlos starten →