Keynote Speakers

Gait recognition aims to identify individuals by their walking style, and is a typical long-range, contact-free biometric technology.
This talk first revisits the fundamental problem setting of gait recognition.

It then introduces recent vision foundation model approaches, represented by BigGait and BiggerGait, and explains how large-scale pretraining improves representation learning and cross-domain generalization, driving gait recognition from task-specific models toward more general identity representation learning.
More broadly, the concept of gait recognition is gradually extending beyond its original scope to a wider problem of identity recognition in videos.

Identity recognition in videos is not limited to walking patterns, but relies on holistic modeling of visual information from the human body.
Human visual identity modeling is a more general research framework for future study supported by large vision models.

Dr. Shiqi Yu is an Associate Professor at the Department of Computer Science and Engineering at the Southern University of Science and Technology (SUSTech).

His main research areas are gait recognition and object detection. In the field of gait recognition, he created the CASIA-B gait database, which is widely used in gait recognition, and the OpenGait open-source project, which has become a major algorithm evaluation framework.

He has published more than 100 papers on gait recognition in IEEE TPAMI, IEEE TIFS, IEEE TBIOM, PR, CVPR, ECCV, IJCB, etc. In terms of object detection, his face detection algorithm was adopted by numerous enterprises.
He has been listed among the Top 2% Scientists (Career Long) since 2025, and Top 2% Scientists (Single Year) since 2021.

Dr. Yu currently serves as a supervisor of China Society of Image and Graphics (2020-2025, 2025-2030), Chair of IAPR TC4 (2025-2026), Vice President of IEEE Biometrics Council (2024-2025, 2026-2027), and a Board Member of the OpenCV Foundation (since 2020).

He has previously served as the program chair of CCBR 2017, CCBR 2021 and IJCB 2024, as well as the general chair of CCBR 2023 and the 14th and 15th Guangdong-HK-Macao Conference on Image and Graphics. Moreover, he has been the main organizer of the IAPR/IEEE Winter School on Biometrics, held annually in Shenzhen since 2018.

Biometrics have become a key technology for identity verification, with face recognition now widely deployed in applications such as border control and remote identity verification.

As these systems become more widespread, they must deliver a seamless and inclusive user experience while maintaining a high level of security. This balance is increasingly difficult to achieve as attacks evolve, from advanced presentation attacks using realistic materials to digital attacks powered by generative AI and deepfake technologies.

In this presentation, I will explain why addressing this challenge requires moving beyond standalone biometric algorithms toward an end-to-end system-level approach. At Thales, we combine accurate biometric algorithms, image quality assessment, sensor control, cybersecurity, and trustworthy AI to build biometric systems that are both user-friendly and resilient against emerging threats.

Sandra Cremer is Head of Biometrics and AI Research at Thales, within the Cybersecurity and Digital Identity division, where she leads research in biometric technologies for secure identity verification.

She holds a PhD in Computer Science from Institut Mines-Télécom and has worked in the field of biometrics throughout her career, evolving from research scientist to leadership roles in charge of Biometric Research & Technology activities. Prior to her current position, she led biometric research teams at IN Groupe and contributed to identity system developments at Thales.

Her expertise spans multiple biometric modalities, including fingerprint, face, and iris recognition, with applications in civil identity, border control, travel, and public security. She has authored scientific publications and patents in areas such as iris recognition and presentation attack detection.

The focus of the teams she currently leads is to address real-world challenges in biometric systems, including robustness to spoofing, morphing, and deepfake-based attacks, and on developing high-performance algorithms for end-to-end systems to deliver secure, reliable, and user-friendly identity solutions.