Past & Upcoming Talks

This presentation outlines the research ecosystem, technological infrastructure, and recent advancements of the LITRP laboratory in the field of large-scale biometric recognition systems. As an internal, contactless feature, vein recognition offers high security, distinctiveness, and resistance to presentation attacks or surface skin variations. To overcome practical limitations such as data scarcity, stringent privacy regulations, and scalability obstacles in evaluating large-scale deep learning models , the research heavily explores the generation of large-scale synthetic datasets. Utilizing both style-based generative architectures (StyleGAN) and nature-inspired methodologies (Physarum polycephalum growth modeling) , the team successfully generated datasets scales up to millions of distinct synthetic palm vein images (e.g., the LS-SPVDB dataset containing 120 million images). Furthermore, the presentation displays state-of-the-art verification rates utilizing advanced feature extraction methodologies, parallel KNN classifiers, convolutional neural networks (CNN), and end-to-end multi-task learning frameworks that integrate identity matching with soft biometrics like age and gender. Finally, current and future directives focus on trustworthy and privacy-preserving biometrics via generative AI to safely replace real data during rigorous algorithm testing.

Some of the projects developed in the Laboratory of Technological Research in Pattern Recognition of the Universidad Católica del Maule were presented in this lecture. Since 2018, our Lab develops different research in the area of biometrics on high-performance computing that aims to the massive recognition of people, with very good results on fingerprint and palm veins as biometric features. Besides, we introduced the results advances of the project that we currently develop with the VIP Research Group for the creation of a multitask learning model for palm vein recognition and a new multilabel database of palm vein images.

Palm vein recognition is an emerging biometric technique with several advantages, especially in terms of security against forgery, which has gained the attention of the research community. However, collecting large-scale biometric datasets is a challenging task because of restrictions on time, security, and cost. Publicly available databases have a reduced number of individuals and samples lacking detailed annotations on soft traits (e.g., gender, age, weight). Therefore, evaluating the scalability of developed methods on massive datasets is not feasible, and the influences of different attributes have been poorly investigated. This talk will dive deeply into the suitability of synthetic vein images generated to compensate for the urgent lack of publicly available large-scale datasets. Firstly, it will present an overview of recent research progress on palm vein recognition, from the basic background knowledge to vein anatomical structure, data acquisition, public database, and recognition approaches. Later, it will also examine state-of-the-art methods that have allowed the generation of vascular structures for biometric purposes. Finally, it will introduce a general flowchart for creating a synthetic palm vein database and a conceptual mathematical model to generate synthetic palm vein images, analyzing the performance of synthetic palm vein datasets for biometric recognition.

Workshop on theoretical foundations of Convolutional Neural Networks (CNN) and their applications. It provides an approach to the components of CNNs, main architectures, procedures, and best practices for CNN training. Additionally, a practical section addresses the tools available for implementing, training, and executing convolutional networks to solve practical cases from end-to-end and pre-trained models using Matlab and Python.

Biometric people identification is an essential mechanism in modern society. Palm vein recognition is an emerging biometric technique with several advantages, especially in terms of security against counterfeiting and stability over time. Moreover, palm vein-based biometric systems are characterized by their contactless data acquisition, which makes them more suitable for several applications in today's world. This talk provides an approach to the state-of-the-art of this novel biometric technique, exploring its main features, results, uses, and challenges. Also, it is presented the results obtained by the Laboratory of Technological Research in Pattern Recognition (LITRP) of the Universidad Católica del Maule.

Lecture on related topics in Supercomputing, AI, and how both are combined to facilitate the processing of large amounts of data and the development of new computing systems to address current technological challenges in biometrics applications.