Past Events

The team of the Department of Multimedia Systems of the Gdańsk University of Technology in Poland has been working with the largest Polish bank in the field of biometrics since 2015. They have constructed an experimental multimodal biometric workstation replicated in 100 copies and installed in 60 branches of this bank. This workstation allowed them to conduct a large experiment in which they investigated the usability of different banking biometric modalities. The biometric system includes a biometric pen the team constructed to verify handwritten signatures, a camera and their proprietary facial recognition software, a microphone that works with an artificial neural network to verify the speaker, and a hand-vein sensor.

The team at GUT involved over 150 bank advisors and nearly 10.000 bank customers in the experiment. Also, they used surveys to gather opinions on the usefulness of the different types of biometrics. They have currently constructed an upgraded testbed, which has been extended with a scanner of ID documents and a gaze tracker, enabling contactless control of the whole biometric authentication process, and it also provides yet another biometric modality based on the analysis of saccadic movements.

In their video footage presentation, Prof. Czyżewski will talk about their construction work, research, results, and work in progress.

Curriculum vitæ

Prof. Andrzej Czyżewski is the Head of the Department of Multimedia Systems of the Gdańsk University of Technology in Poland. His scientific interests focus on multimedia techniques and their numerous interdisciplinary applications for telemedicine, traffic control, smart cities, biometrics, and others.

Prof. Czyżewski lectures at the Gdańsk University of Technology on "Multimedia technologies" and "Sound and image engineering." He cooperates with, among others with the Medical University of Gdańsk and with some large international corporations, with the PKO Bank, with the General Directorate of National Roads and Motorways, and with many smaller enterprises in which he and his team implement inventions and innovations being developed. He is the author of over 600 scientific publications and nearly 30 patents. He has so far promoted 13 doctors (Ph.D.) and over 300 masters of science (M.Sc.) and engineers.

This talk will cover the iris presentation attack detection. With a backbone based on two survey papers on this topic recently co-authored by Czajka (published in ACM Computing Surveys, 2018 and in Pattern Recognition Letters, 2020), Czajka will primarily discuss the problem of recognition of unknown attack types (unseen during training), or less popular attack types (such as cadaver eyes), and will bring a few ideas of making the iris PAD algorithms more attack type-agnostic. A short summary of available iris PAD datasets, open-source iris PAD solutions, and the performance in an open-set classification regime observed in the fourth edition of LivDet-Iris competition (concluded in 2020) will be delivered.

Curriculum vitæ

Adam Czajka is an Assistant Professor in the Department of Computer Science and Engineering at the University of Notre Dame. His research focuses on biometrics and security, especially on human-driven methods of detecting unknown types of presentation attacks driving a biometric system into incorrect decision. In general, he is fascinated by wide spectrum of research topics in computer vision, pattern recognition and machine learning. Recently, he has been exploring the use of visual signals generated by artificial neural networks in stimulation of emotions and mental healing processes. In the classroom, he enjoys teaching graduate- and undergraduate-level Computer Vision (CSE 40535/60535/40536/60536), Biometrics (CSE 40537/60537) and Neural Networks (CSE 40868/60868) courses.

Czajka is originally from Poland. He received M.Sc. in Computer Control Systems (hons), Ph.D. in Biometrics (hons) and D.Sc. degree ("habilitation") in Computer Science, all from Warsaw University of Technology (WUT). Before coming to Notre Dame, he was the Chair and co-founder of the Biometrics and Machine Learning Laboratory in the Institute of Control and Computation Engineering at WUT, the Head of the Postgraduate Studies on Security and Biometrics, the Vice Chair of the NASK Biometrics Laboratory, and the Chair of the Polish Standardization Committee on Biometrics. He is a Senior Member of the IEEE, an active member of the EAB, an associate member of the IAI and an expert of ISO/IEC Sub-Committee 37 on Biometrics.

The xAI4Biometrics Workshop aims at promoting a better understanding, through explainability and interpretability, of currently common and accepted practices in several and varied applications of biometrics. These applications, in scenarios comprising identity verification for access/border control, watching lists surveillance, anti-spoofing measures embedded in biometric recognition systems, forensic applications, among many others, affect the daily life of an ever-growing population.

The xAI4Biometrics Workshop @ WACV 2021 will include five oral paper presentations and two keynote talks:

• "Explaining Deep Neural Network Decisions for Enhanced Biometric Systems" by Prof. Peter Eisert (Humboldt University Berlin and Fraunhofer HH)
• "At the Two Extremes of Interpretability in Machine Learning: Very Sparse Models and Interpretable Deep Neural Networks" by Prof. Cynthia Rudin (Duke University).

You are still on time to register and remotely attend, not only the workshop but also the WACV2021 conference for a very affordable fee.