Research seminars

Abstract: Generative AI—so-called deepfakes—have captured the imagination of some and struck fear in others. Although they vary in their form and creation, deep fakes refer to text, image, audio, or video that has been automatically synthesized by a machine-learning system. Deepfakes are the latest in a long line of techniques used to manipulate reality, yet their introduction poses new opportunities and risks due to the democratized access to what would have historically been the purview of Hollywood-style studios. I will describe how synthetic media is created, how it is being used and misused, and if (and how) it can be perceptually and forensically distinguished from reality.

Bio: I am a Professor at the University of California, Berkeley with a joint appointment in Electrical Engineering & Computer Sciences and the School of Information. My research focuses on digital forensics, forensic science, misinformation, image analysis, and human perception. I received my undergraduate degree in Computer Science and Applied Mathematics from the University of Rochester in 1989, and my Ph.D. in Computer Science from the University of Pennsylvania in 1997. Following a two-year post-doctoral fellowship in Brain and Cognitive Sciences at MIT, I joined the faculty at Dartmouth College in 1999 where I remained until 2019. I am the recipient of an Alfred P. Sloan Fellowship, a John Simon Guggenheim Fellowship, and am a Fellow of the National Academy of Inventors.

Abstract: Mobile robots should be aware of their situation, comprising the deep understanding of their surrounding environment along with the estimation of their own state, to successfully make intelligent decisions and execute tasks autonomously in real environments. 3D scene graphs are an emerging field of research modeling the environment with high-level semantic abstractions (such as chairs, tables, or walls) and their relationships (such as a set of walls forming a room or a corridor). While providing a rich understanding of the scene, they typically rely on separate SLAM methods that previously estimate the robot’s pose and its map using metric/semantic representations without exploiting this hierarchical high-level information. Thus, in general, 3D scene graphs and SLAM graphs are loosely coupled. In our work named Situational Graphs, we propose for the first time a tight coupling of geometric LiDAR SLAM with 3D scene graphs in a single optimizable factor graph, demonstrating state-of-the-art metrics. Furthermore, we extend Situational Graphs to showcase the strength of hierarchical mapping for localizing a robot using LIDAR data and architectural plans. We have also demonstrated improved results when extending Situational Graphs to pure vision-based systems such as RGB/RGB-D. Finally, Situational Graphs have also been exploited for fast and robust hierarchical high-level planning in complex indoor environments. 

Bio: Hriday Bavle received his PhD in Automatic Control and Robotics from the Technical University of Madrid, Spain in 2019. His PhD was focused on improving localization and mapping algorithms on-board aerial robotic platforms for augmenting their awareness and autonomy in cluttered indoor environments. During his PhD, he also served as a visiting researcher at the Aerospace Controls Lab of Massachusetts Institute of Technology, Boston, USA for 3 months where he developed a semantic SLAM algorithm running on-board an aerial robot flying at high speeds. After his PhD, he worked as a robotics researcher in a robotics startup in Denmark. Hriday is currently a postdoctoral research associate at the University of Luxembourg since 2021 researching novel techniques coupling SLAM graphs and 3D scene graphs with a final goal of providing mobile robots with situational awareness similar to humans. 

Abstract: The recent explosive progress in text-to-image generation has taken both the research community and the general public by surprise. These impressive capabilities were unlocked using new image generation mechanisms and large language models -- a combination that appears to scale and improve limitlessly when fed with ever more training data and compute. While these methods let us sidestep many difficulties in classical image synthesis -- like building geometry, and simulating lighting and materials -- we are now at the mercy of the poorly understood behavior of the models, with rather blunt tools for controlling the outcomes. In this talk, I'll trace the origins of this sudden "Big Bang" event, and introduce the key historical and current players in the field, namely GANs, denoising diffusion, and language models. I'll also discuss their controllability, and place in the broader space of content authoring.

Bio: Miika Aittala is a Senior Research Scientist at NVIDIA Research, which he joined in 2019. He received his PhD in 2016 from Aalto University, working on material appearance capture and rendering algorithms. Prior to his current position, he worked as a postdoctoral researcher at MIT CSAIL and visited Inria Sophia Antipolis. His current research focuses on neural generative modeling.

Abstract: Have you ever wondered what it's like to make the move from Zaragoza to the United States? Or what the journey of founding a company in the Aerospace Industry entails? In this presentation, Rocio will share her personal experience as Founder and CEO, offering insights into the challenges and triumphs of the past few years. At the core of Improving Aviation is a profound commitment to fostering a more equitable, sustainable, and intelligent world through innovative technological solutions.

Bio: Rocio Frej Vitalle is the Founder and CEO of Improving Aviation LLC, an aerospace engineer with over two decades of experience in the industry. Rocio holds a B.S. in Aerospace Engineering from Embry-Riddle Aeronautical University and a Master’s Degree in Civil Engineering & Transportation from the University of South Florida.

Improving Aviation is dedicated to advancing aerospace technologies for enhanced environmental sustainability and disaster management in the face of climate change. Our expertise in aerospace engineering extends to providing services for both private and federal organizations, including esteemed institutions like the Federal Aviation Administration, NASA, NOAA, and the Air Force.

Our commitment to excellence has been recognized with numerous accolades, including being named the winner of the 2021 TERP Business Expo in Daytona Beach and Startup of the Year by Synapse, Florida's innovation community connector. We take pride in contributing to the industry's growth, with Rocio serving on the Young Aviation Professionals Committee at the Air Traffic Control Association (ATCA) and as a session chair at the American Institute of Aeronautics and Astronautics (AIAA). Rocio was also honored with the 'Article of the Year' award by the Air Traffic Control Journal.

At Improving Aviation, we are not just shaping the future of aerospace; we are actively involved in making it more sustainable, resilient, and innovative. Our dedication to these principles is evident in every project we undertake. 

Abstract: This talk is about robotic manipulation and perception of canonical "deformable linear objects" like a Kirchhoff elastic rod (e.g., a flexible wire). I continue to be amazed by how much can be gained by looking carefully at the mechanics of these objects and at the underlying mathematics. For example, did you know that the free configuration space of an elastic rod is path-connected? I'll prove it, and tell you why it matters.

Bio: Timothy Bretl comes from the University of Illinois at Urbana-Champaign, where he is a Professor, a Severns Faculty Scholar, and the Associate Head for Undergraduate Programs in the Department of Aerospace Engineering. He holds an affiliate appointment in the Coordinated Science Laboratory, where he leads a research group that works on a diverse set of projects in robotics and education (http://bretl.csl.illinois.edu/). He has also received every award for undergraduate teaching that is granted by his department, college, and campus.

Abstract: Animation plays a crucial role in having engaging virtual humans in VR. Whether we deal with a group of autonomous agents whose behavior will increase the sense of presence by making the environment feel more plausible, or with the movement of our own avatar whose movement quality will have an impact on embodiment and task performance. In this talk I will be covering our most recent work on studying the impact of animations on self avatars with the purpose of developing new animation models that can be driven by sparse input data, as well as animating autonomous agents to improve the fidelity of animation impact on the environment, as well as studying the influence that animation quality can have on presence and user experience.

Bio: Nuria Pelechano is an Associate Professor at the Universitat Politecnica de Catalunya, Spain, where she is a member of the Research Center for Visualization, Virtual Reality and Graphics Interaction (ViRVIG). She is the president of the Eurographics Spanish Chapter, and member of the Sociedad Científica Española de Informática (SCIE). She has co-authored two books on Crowd Simulation and has published in journals and conferences on computer graphics and animation (Transactions on Graphics, Computer Graphics Forum, Computers & Graphics). She has participated in projects funded by the EU (currently MSCA ITN-ETN CLIPE, HORIZON-101093159-XR4ED), Spanish Government (currently SENDA: TED2021-129761B-I00, PID2021-122136OB-C21), and USA institutions, and also worked on technology transfer projects regarding crowd evacuation, and applications of Virtual Reality for architecture design. Her research interests include simulation, animation and rendering of crowds, generation of navigation meshes, and studying human-avatar interaction in Virtual Reality.

Abstract: In a classical deep learning computer vision pipeline, a huge amount of human efoort is put in the collection and annotation of image data. This is necessay to train well performing neural networks. Especially for industrial computer vision applications, where the need for high accuracies is paramount, the costs of data collection and annotation  are skyrocketing. Therefore, in this talk, prof. Toon Goedemé will give an overview of techniques that emeliorate the need for large amounts of (annotation) data in the training of custom computer vision networks. The talk will be illustrated with multiple real-life industrial application examples that were studied at prof. Goedemé's EAVISE lab at KU Leuven, Belgium.

Bio: Toon Goedemé is a professor at KU Leuven (Belgium), where he conducts research into applied computer vision and AI. He is also the founder and research leader of EAVISE, a research group dedicated to developing innovative algorithms and applications for image processing and analysis. Goedemé obtained his PhD in engineering sciences from KU Leuven in 2006, with a thesis on real-time visual navigation for mobile robots, under guidance of prof. Luc Van Gool and prof. Tinne Tuytelaars. Since then, he founded EAVISE as a part of the ESAT-PSI group. The research goal of EAVISE is applying state-of-the-art techniques from computer vision and artificial intelligence as a solution for industry-specific problems. In order to meet stringent execution speed, energy footprint, cost and price requirements, the developed algorithms are implemented and optimized on embedded systems, such as GPUs, DSPs and FPGAs. Application areas include industrial automation, product inspection, traffic monitoring, e-health, agriculture, eye-tracking research, microscopic imaging, surveillance and land surveying.

Abstract: There have been three revolutions in human computing: Personal computers in the 1980s, the World Wide Web and cloud computing in the 1990s, and the iPhone in 2007. The fourth one will include augmented and virtual reality (AR/VR) technology. The ability to transfer human motion from AR/VR sensors to avatars will be critical for AR/VR social platforms, video games, new communication systems, and future workspaces. In the first part of this talk, I will describe several techniques to compute subtle human behavior with applications to AR/VR (e.g., facial expression transfer to photorealistic avatars)  and medical monitoring/diagnosis (e.g., depression diagnosis from audio/video). In addition, I will show how we can estimate dense human correspondence from WiFi signals,  which could pave the way for novel AR/VR interfaces. 

All these techniques for human sensing rely on training deep learning models. However, in practice metric analysis on a specific train and test dataset does not guarantee reliable or fair ML models. This is partially due to the fact that obtaining a balanced (i.e., uniformly sampled over all the important attributes), diverse, and perfectly labeled test dataset is typically expensive, time-consuming, and error-prone.  In the second part of this presentation, I will introduce two methods aimed at enhancing the robustness and fairness of deep learning techniques.  First, I will delve into a technique for conducting zero-shot model diagnosis. This technique allows for the assessment of failures in deep learning models in an unsupervised manner, eliminating the need for test data. Additionally, I will discuss a method designed to rectify biases in generative models, which can be achieved using only a small number of sample images that showcase specific attributes of interest.

Bio: Fernando De la Torre received his B.Sc. degree in Telecommunications, as well as his M.Sc. and Ph. D degrees in Electronic Engineering from La Salle School of Engineering at Ramon Llull University, Barcelona, Spain in 1994, 1996, and 2002, respectively. He has been a research faculty member in the Robotics Institute at Carnegie Mellon University since 2005. In 2014 he founded FacioMetrics LLC to license technology for facial image analysis (acquired by Facebook in 2016). His research interests are in the fields of Computer Vision and Machine Learning. In particular, applications to human health, augmented reality, virtual reality, and methods that focus on the data (not the model). He is directing the Human Sensing Laboratory (HSL).

Abstract: The increasing adoption of Artificial Intelligence (AI) systems in high-stakes applications brings new opportunities for innovation, economic growth and the digital transformation of society. However, this often comes with associated risks to the safety, health or fundamental rights of people, highlighting an urgent need for the systematic adoption of trustworthy AI practices. Trustworthy AI is still in its infancy, although the topic is widely recognized as one of the major research priorities in the field. Both key academic and industrial players have recently proposed good practices towards trustworthy AI, including methodologies for improving AI systems' transparency, explainability, fairness and non-discrimination, among others.

At this moment, a transition from voluntary good practices to hard legal requirements is underway with the adoption of AI regulatory frameworks appearing on the horizon in many parts of the world. In particular, trustworthy AI is at the heart of the new European AI regulation proposal, the AI Act. In this seminar, we will dive into the AI Act, analyze its requirements, its human-centred focus and the transformative technical, organizational and social changes it will bring. We will explore the current research landscape on trustworthy AI methodologies, related challenges and benefits. We will illustrate trustworthy AI and AI Act's principles through the example of real-world facial processing applications, recommender systems and generative AI.

Bio: Isabelle Hupont Torres holds a PhD from the University of Zaragoza and is a researcher in Artificial Intelligence (AI), with more than 17 years of experience in the field. Her research has been particularly focused on affective computing, facial analysis and human-machine interaction. Isabelle has more than 80 scientific publications in prestigious international journals and conferences, and has participated in more than 35 projects in public, national and European research and innovation projects. 

She is currently a Scientific Officer at the Joint Research Center of the European Commission, where she explores the regulatory side of technology and contributes to setting the path towards trusted AI and digital worlds in the European Union. Isabelle is strongly committed to women's rights, actively encouraging young women to take an interest in technical careers, and advocating for the need to apply an inclusive, fair and diverse perspective in AI. 

Abstract: Robotics plays a pivotal role in the strategy to expand the competitiveness of all sectors of the economy, as well as offering new solutions to societal challenges. While industrial robots typically perform repetitive tasks, operating in industrial settings where the world is designed around them, service robots assist humans at work in non-industrial settings or in the home, needing to operate in complex dynamic unstructured environments. Robots need to continuously acquire a complete situational awareness in such environments (i.e., perceiving the environment within time and space, comprehending its meaning, and projecting it in the future) to enable intelligent decision-making and autonomous task execution. Although the robotics research community has been investigating topics such as perception, localization, and mapping for many years, there is a strong need for more advanced concepts of robotic situational awareness as an essential core robot’s capability. Therefore, the Automation and Robotics Research Group at the Interdisciplinary Centre for Security, Reliability, and Trust (SnT), University of Luxembourg, has a major focus on such situationally-aware robots.

In this talk, I will present our latest research featuring a novel concept to provide robots with a deeper understanding of the situation with multiple levels of abstraction such as geometric (e.g., shape of the objects), semantic (e.g., type of the objects), or relational (e.g., relationships between the objects). This understanding is not limited to the current situation (i.e., what is the situation that the robot perceives at the present moment), but it creates a long-term situational understanding (i.e., what is going on around the robot) by integrating the perceived measurements from multiple sensor sources with past observations and background knowledge. This is achieved by combining adapted versions of novel machine learning based techniques with extended versions of traditional optimization-based algorithms, to get the best out of each of them. Finally, I will present a decision-making algorithm (concrete for path planning) that exploits previously mentioned advanced situational awareness to obtain more efficient and more precise results.

Bio: Dr. Jose Luis SANCHEZ-LOPEZ is a Research Scientist on permanent position at SnT in the University of Luxembourg (UL) since 2021. He is the co-head of the Automation and Robotics Research Group (ARG), and the head of the Mobile Robotics Lab. In his more than 12 years of experience, he has worked on providing robots with essential capabilities to make them more autonomous and intelligent. His contributions are in three main areas: perception and situational awareness, intelligent and cognitive system architectures, and trajectory and path planning and control. All the outcomes of his research have always been experimentally validated in applied research and technology transfer projects, being the scientific results disseminated in more than 85 publications in renowned international peer-reviewed journals and conference proceedings, with an impact of around 2100 citations (according to Google Scholar).

Abstract: Robot swarms have the potential to assist us in tasks involving vast scenarios, robots in persistent (24/7) missions with added resilience so that they can complete their objectives despite unforeseen difficulties, and simpler logistics. However, current demonstrations of swarm technology in unstructured environments only count on single-digit individuals. That is farther from what one would expect from the huge scaling potential of a swarm. In this talk, I will present some practical challenges that mobile robot swarms face in fundamental tasks, e.g., the control of specific geometry parameters during the deployment of a mobile robot swarm, also known as formation control. As an application of higher-level tasks leveraging formation control, we will see the coordination of robots while tracking paths and the source-seeking of scalar fields. Onboard imperfections are responsible for non-designed emergent behavior and might stop us from the desired scalability of the system; nevertheless, hidden opportunities within the imperfections could assist us with practical deployments

Bio: Héctor is a Ramón y Cajal researcher in the Computer Engineering, Automation and Robotics (ICAR) at the University of Granada where he runs the Swarm Systems Laboratory. He is the recipient of an ERC Starting Grant where he investigates the impact of imperfections on robot swarms, associate editor of the IEEE Transactions of Robotics, and proud developer of the open-source drone project Paparazzi. He received his Ph.D. in systems and control from the University of Groningen, the Netherlands, in 2016. Afterward, he was a postdoctoral research associate with the Ecole Nationale de l’aviation Civile, Toulouse, France, and an assistant professor with the Unmanned Aerial Systems Center, University of Southern Denmark, Odense, Denmark.

Abstract: Exploration of large, complex virtual environments is an integral part of an immersive experience in virtual reality (VR). However, safe exploration of virtual environments is difficult since the virtual world is generally much larger than the user's physical environment, meaning that an unobstructed path in the virtual world may correspond to an obstructed path in the physical world. Locomotion interfaces are techniques that allow users to move through virtual environments without colliding with physical objects. Existing interfaces, such as walking-in-place or teleportation, can enable long-distance exploration, but are not immersive due to their unrealistic controls. Interfaces that let users navigate using natural, everyday walking are usually preferred since they are more intuitive and create a higher sense of presence, but they usually require a large physical space in order to be used effectively.

In this talk, I will discuss how we combine techniques from human visual perception and robot motion planning to develop new algorithms that enable users to explore large virtual environments using natural walking, with a focus on interfaces that can function outside of controlled lab environments. To achieve this, we leverage the concept of alignment to develop new locomotion interfaces that take into account the structure of both the physical and virtual environment to optimally steer the user away from physical obstacles that they cannot see. Additionally, we introduce new algorithms that are built using a formalization of the VR locomotion problem based on motion planning. Our mathematical formalization allows us to leverage techniques from robot motion planning and computational geometry to develop steering algorithms that are more easily generalizable to different environment layouts without requiring us to change the algorithm implementation. Finally, I will present ongoing work that leverages interactable components of the virtual environment to further influence users to travel on collision-free paths while providing an immersive virtual experience.

Bio: Niall is a PhD student at the University of Maryland where he is advised by Prof. Dinesh Manocha and Prof. Ming Lin. His research interests include virtual/augmented reality, human perception, computer graphics, and robotics. Niall has interned at Meta Reality Labs and NVIDIA Research, and received his bachelor's degree in computer science from Davidson College.

Abstract: Virtual reality (VR) is an emerging medium that has the potential to unlock unprecedented experiences. Since the late 1960s, this technology has advanced steadily, and can nowadays be a gateway to a completely different world. VR offers a degree of realism, immersion, and engagement never seen before, and lately we have witnessed how newer virtual content is being continuously created. However, to get the most out of this promising medium, there is still much to learn about people’s visual attention and gaze behavior in the virtual universe. Questions like “What attracts users’ attention?” or “How malleable is the human brain when in a virtual experience?” have no definite answer yet. We argue that it is important to build a principled understanding of viewing and attentional behavior in VR. This thesis presents contributions in two key aspects: Understanding and modeling users’ gaze behavior, and leveraging imperceptible manipulations to improve the virtual experience.

In the first part of this thesis we have focused on developing computational models of gaze behavior in virtual environments. First, we have devised models of user attention in 360º images and 360º videos that are able to predict which parts of a virtual scene are more likely to draw viewers’ attention. Then, we have designed another two computational models for spatio-temporal attention prediction, one of them able to simulate thousands of virtual observers per second by generating realistic sequences of gaze points in 360º images, and the other one predicting different, yet plausible sequences of fixations on traditional images. Additionally, we have explored how attention works in 3D meshes. The second part of this thesis attempts to improve virtual experiences by means of imperceptible manipulations. We have firstly focused on lateral movement in VR, and have devised thresholds for the detection of such manipulations, which we then applied in three key problems in VR that have no definite solution yet, namely 6-DoF viewing of 3-DoF content, overcoming physical space constraints, and reducing motion sickness. On the other hand, we have explored the manipulation of the virtual scene, resorting to the phenomenon of change blindness, and have derived insights and guidelines on how to elicit or avoid such an effect, and how human brains’ limitations affect it.

Bio: Dani got his BSc and MSc in Computer Science at the Universidad de Zaragoza, and is currently a last-year PhD student in the Graphics and Imaging Lab, under the supervision of Prof. Belen Masia and Prof. Diego Gutierrez. His research mainly spans virtual reality, and encompasses topics such as understanding and modeling visual attention and gaze behavior, multimodality, content generation, or studying diverse perceptual manipulations. During his PhD, he has done two research stays at Adobe Research, one under the supervision of Dr. Xin Sun, and another one supervised by Dr. Aaron Hertzmann and Dr. Stephen DiVerdi, and one research stay at Meta Reality Labs Research, supervised by Dr. Michael Proulx. He was also granted a Fulbright Predoctoral Scholarship to conduct his research in the US for six months. More details can be found at https://webdiis.unizar.es/~danims/.

Bio: Alex Ginés is a Growth Chief Officer at Nakima and Immersive Technologies technician at Neàpolis. He has been working on immersive technological projects since 2015 and collaborating with different research centres and universities such as: UPC ESPEVG, UPF, URV, I2CAT, LEITAT among others. Nakima as an innovation factory, his goal is to help third companies to squeeze their ideas and projects to reach their own goals.

Abstract: In today's dynamic entrepreneurial landscape, the journey from concept to creation is both exhilarating and challenging. This workshop delves into the transformative process of bringing ideas to life, whether they manifest as innovative products or groundbreaking concepts ripe for spinoff ventures. Throughout the session, participants will explore key strategies, practical insights, and real-world examples that illuminate the pathway to entrepreneurial success.

The workshop begins by unraveling the anatomy of a compelling idea, dissecting the essential elements that distinguish mere concepts from potential game-changers. Participants will learn how to cultivate creativity, identify market opportunities, and leverage emerging trends to generate ideas with transformative potential.

Bio: Dorian obtained his PhD on computer vision for localization in SLAM in 2013, in the University of Zaragoza, Spain. He was a postdoctoral researcher at the University of Zaragoza and also at the Autonomous Robotics & Perception Lab at the University of Colorado Boulder, and The George Washington University (DC) in USA. 

Dorian moved from academia to industry in 2015 to work as a Senior Computer Vision Researcher at Paracosm, a start-up based in Florida. We launched the hand-held device PX-80 to perform 3D reconstruction of large environments with fisheye, lidar and IMU sensors. After that, I joined Intel RealSense, where we released the Intel® RealSense™ Tracking Camera T265 in 2019, a SLAM-on-a-chip device. Dorian is currently Senior Computer Vision Engineer at Niantic. His main interests are on Computer vision and Software engineering, in particular on the topics of Real-time localization for visual SLAM and Object recognition for augmented reality.

Abstract: Visual Positioning Systems (VPS) combine Simultaneous Localization And Mapping (SLAM) and 3D Reconstruction to offer geolocalization as a service to developers of augmented reality experiences, video games, mobile vehicles, etc., on camera-equipped devices. In addition to the local positioning provided by typical visual SLAM systems, a global VPS can also accurately locate the user in the world by integrating the coverage of GPS sensors with the precision of SLAM systems.

This talk provides an overview of the technology and algorithms that support VPS systems and explores the functionalities that platforms in the market offer to developers, enabling them to build their solutions without requiring an in-depth understanding of computer vision.

Bio: Antonino Furnari is a tenure-track assistant professor at the University of Catania. He received a PhD degree in Mathematics and Computer Science in 2017 from the same institution and spent time as a visiting researcher at the University of Texas at Austin and at the University of Bristol. He has been working on First Person (Egocentric) Computer Vision since 2014 and he is part of the EPIC-KITCHENS and EGO4D teams. His research focuses on understanding human activity and future intent from egocentric video.

Abstract: Computer Vision and Artificial Intelligence have seen a surge in popularity in the last years thanks to important technological and algorithmic breakthroughs. While the public starts confronting with the paradigm shift involved in the widespread adoption of such new technology, access to AI is still limited to few individuals and mediated through devices, such as computers and smartphones, which detach humans from the reality. To democratize access to AI, we need to build systems which can perceive the world as humans do and swiftly provide them feedback while they are still engaged in their activities, without requiring any context switch. In this talk, we'll see how egocentric vision, i.e., the analysis of images, videos and multimodal data captured from the user's point of view, can answer this call. We'll discuss the challenges and opportunities provided by the field, point out the scientific contributions and the technological tools which make research in this field possible, showcase the main applications, and highlight the open research questions.

Bio: Rosa obtained her bachelor and Master at the University of Zaragoza. Since 2017 she works at ATRIA Innovation, where she currently leads the Industry 4.0 projects, working on improving the efficiency, safety and quality of numerous production processes, applying computer vision, augmented reality or robotics solutions.
Jorge holds a B.S. in physics and a Master's Degree in Robotics, Graphics and Computer Vision both from the University of Zaragoza. After finishing his studies he obtained an internship at ATRIA Innovation and is now a full-time materials and sustainability engineer.

Abstract: Innovation is closer than you think! ATRIA, right here in Zaragoza helps factories run smoother and smarter, all while building a sustainable future. 

Conventional RGB cameras capture a limited portion of the electromagnetic spectrum. Hyperspectral imaging, however, analyses the unique spectral fingerprint of each pixel within an image. This allows for the identification of materials and real-time monitoring of complex processes. This talk will cover two real projects developed recently at ATRIA and relate them with the RGCV master's subjects. Both projects involved hyperspectral data acquisition, measurement, and the training of machine learning models to address specific industrial challenges. These projects show examples to bridge the gap between theoretical knowledge gained in a RGCV Master's program and its practical application in the field.