The CarMen consortium is proud to share a significant scientific achievement from project partner IDIAP Research Institute.
A new research paper entitled “Lightweight Cross-Spectral Face Recognition via Contrastive Alignment and Distillation” has been accepted for publication in the prestigious IEEE Transactions on Biometrics, Behavior, and Identity Science (TBIOM) journal.
Developed by researchers from the Biometrics Security and Privacy Group at IDIAP Research Institute, the paper presents an innovative and lightweight framework for Heterogeneous Face Recognition (HFR), enabling accurate matching between facial images captured under different spectral conditions, including visible light (RGB) and near-infrared (NIR) imaging.
The proposed approach adapts a hybrid CNN-Transformer architecture originally developed for RGB face recognition and enhances it through contrastive alignment and knowledge distillation techniques. This enables efficient end-to-end training using only a limited amount of paired heterogeneous data while maintaining strong performance on standard face recognition benchmarks.
This achievement directly supports the objectives of the CarMen project, which aims to develop next-generation biometric solutions for seamless and secure border management. One of the project’s key challenges is ensuring reliable biometric recognition under diverse environmental and operational conditions. As outlined in the project concept, CarMen combines RGB and NIR biometric imaging technologies to improve identification accuracy in real-world border crossing scenarios.
The publication represents an important contribution to the scientific foundations of future biometric systems capable of operating in unconstrained environments while ensuring security, robustness, and privacy protection.
The research was conducted within the framework of the CarMen Horizon Europe project with the support of the Swiss State Secretariat for Education, Research and Innovation (SERI).
Congratulations to Dr. Anjith George and the entire research team at IDIAP Research Institute on this outstanding achievement!