TBD

Dr Ioannis Patras, Professor on Computer Vision and Machine Learning, Queen Mary University of London. Head of Research, Tavus Inc

Abstract:  (TBD)

Bio:

Ioannis Patras is a Professor in the School of Electronic Engineering and Computer Science of Queen Mary, University of London and the Head of Research in Tavus Inc. His research expertise in the area of Multimodal AI, using Machine Learning, Computer Vision and Signal Processing methodologies to learn from multiple sources so as to enable Intelligent Systems to understand the world and humans. Recent works and research interests evolve around

• Learning within and across modalities and in particular learning/adaptation of Large Multimodal Models.
• Learning to recognize and synthesize human behavior and emotions in images, video.
• Large models for controllable generation of Multimodal data, including images and video.

In Tavus, as a Head of Research, is designing Digital Humans using frontier AI models -- interactive agents with voice, face and (emotional) intelligence.

He has published more than 300 articles in the most selective journals and conferences in the area of Computer Vision and Machine Learning, including CVPR, ICLR, NeurIPS, ICCV, ECCV, EMNLP, TPAMI, IJCV, ECCV, TIP. He has been area chair in all of the major conferences in the area of Computer Vision and associate editor in three journals (Image and Vision Computing; Pattern Recognition; and Computer Vision and Image Understanding).  He is a Senior Member of the IEEE and a Fellow of the European Laboratory for Learning and Intelligent Systems (ELLIS).

Dr Dima Damen, Professor of Computer Vision, University of Bristol, UK

Abstract: (TBD)

Bio: 
Dima Damen is a Professor of Computer Vision at the University of Bristol and Senior Research Scientist at Google DeepMind. Dima is currently an EPSRC Fellow (2020-2026), focusing her research interests in the automatic understanding of object interactions, actions and activities using wearable visual (and depth) sensors. She is best known for her leading works in Egocentric Vision, and has also contributed to novel research questions including mono-to-3D, video object segmentation, assessing action completion, domain adaptation, skill/expertise determination from video sequences, discovering task-relevant objects, dual-domain and dual-time learning as well as multi-modal fusion using vision, audio and language. She is the project lead for EPIC-KITCHENS, the seminal dataset in egocentric vision, with accompanying open challenges and follow-up works: EPIC-Sounds, VISOR and EPIC Fields, as well as the recent HD-EPIC. She is part of the large-scale consortium effort Ego4D and Ego-Exo4D. She is an ELLIS Fellow, associate editor (AE) of IJCV, and was a program chair for ICCV 2021 and Associate Editor-in-Chief (AEIC) of IEEE TPAMI (2023-2025). She is frequently a Senior Area Chair and Area Chair in major conferences and was selected as Outstanding SAC in ECCV 2024 and Outstanding Reviewer in CVPR2021, CVPR2020, ICCV2017, CVPR2013 and CVPR2012. Dima received her PhD from the University of Leeds (2009), joined the University of Bristol as a Postdoctoral Researcher (2010-2012), Assistant Professor (2013-2018), Associate Professor (2018-2021) and was appointed as chair in August 2021. She supervises 9 PhD students and 3 Visiting PhD students.

Dr Joost van de Weijer, PI of the Learning and Machine Perception (LAMP) team at the Computer Vision Center (CVC/UAB)

Abstract: 

Model merging has emerged as a promising paradigm for combining multiple task-specific models into a single multi-task model without requiring additional training. In this talk, I will present two recent directions that address key challenges in model merging. First, I will discuss isotropic model merging, which analyzes the role of singular component alignment in successful merging and introduces common and task-specific subspaces to improve performance and reduce the gap to single-task models. Second, I will present Core Space merging, an efficient framework for merging LoRA-adapted models directly in a shared low-rank representation, preserving the computational benefits of parameter-efficient adaptation while improving accuracy. 

Bio: 

Joost van de Weijer leads the Learning and Machine Perception (LAMP) team at the Computer Vision Center, Universitat Autònoma de Barcelona. He received his PhD from the University of Amsterdam in 2005, focusing on physics-based computer vision for improved color understanding in images. He was a Marie Curie Intra-European Fellow at INRIA Rhône-Alpes, France, and later held a Ramón y Cajal fellowship at the Universitat Autònoma de Barcelona. His current research centers on lifelong learning in deep neural networks, encompassing areas such as continual learning, transfer learning, and domain adaptation. In addition, he works on generative models for visual data, with recent efforts aimed at improving efficiency and text-based control in diffusion models. He is a Fellow of the ELLIS Society.

Dr Jai Bardhan, Researcher, Czech Institute of Informatics, Robotics and Cybernetics (CIIRC CTU) 

Abstract: 

Robot world models promise a data-driven replacement for hand-built simulators, supporting policy evaluation, improvement, and planning from video alone. Pretrained video diffusion models are a natural starting point — but a strong visual prior is not yet a usable simulator. This talk presents two ways to adapt one, Stable Video Diffusion, into a deployable world model, both evaluated on DROID. PersistWorld tackles temporal stability: it post-trains the model on its own autoregressive rollouts via RL, closing the loop that otherwise makes long-horizon predictions degrade. DepthWorld tackles geometric grounding: it jointly predicts RGB and depth so rollouts compose into a consistent 3D world, with depth supervision improving RGB prediction itself.

Bio: 

Jai Bardhan received his B.Tech. degree in computer science and M.S. degree in computational natural sciences from IIIT Hyderabad, India, in 2023. He is currently a researcher in the Intelligent Machine Perception group at the Czech Institute of Informatics, Robotics and Cybernetics (CIIRC CTU), Prague, Czech Republic. His current research interests include world models for robotic manipulation, 3D geometric learning, vision language action (VLA) models and self-supervised representation learning.

Dr Vittorio Murino, Professor at the University of Verona &  PI of the AI for Good research unit at the Italian Institute of Technology

Abstract: 

This talk will present a novel few-shot anomaly detection approach, leveraging the pre-trained CLIP model for medical data, and adapting it for both image-level anomaly classification (AC) and pixel-level anomaly segmentation (AS). A dual-branch design is proposed to separately capture normal and abnormal features through learnable adapters in the CLIP vision encoder. To improve semantic alignment, learnable text prompts are employed to link visual features. Furthermore, SigLIP loss is applied to effectively handle the many-to-one relationship between images and unpaired text prompts, showcasing its adaptation in the medical field for the first time. Our approach is validated on multiple modalities, demonstrating superior performance over existing methods for AC and AS, in both same-dataset and cross-dataset evaluations. 

Bio: 

Vittorio Murino is full professor of Computer Vision and Machine Learning at the University of Verona, Italy and principal investigator of the AIGO – AI for Good research unit at the Italian Institute of Technology in Genova, Italy. 

He has been chairman of the Department of Computer Science @ UNIVR from 2001, year of foundation, to 2007. From 2009 to 2019, he worked at the Italian Institute of Technology as principal investigator of Pattern Analysis and Computer Vision (PAVIS) research unit. From 2019 to 2021, he worked as Senior Video Intelligence Expert at the Ireland Research Centre of Huawei Technologies (Ireland) Co., Ltd. in Dublin.

His main research interests are computer vision, machine learning and pattern recognition, specifically focusing on deep learning approaches in imperfect data regime (unsupervised, self-supervised, few/noisy labeling, and biased scenarios), domain adaptation and generalization, and multimodal learning for behavior analysis and related applications.

Prof. Murino is co-author of more than 400 papers published in refereed journals and international conferences, and member of the technical and organization committees of important conferences (CVPR, ICCV, ECCV, ICLR, ICML, NeurIPS, ICPR).

Finally, prof. Murino is IEEE Fellow, IAPR Fellow, and ELLIS Fellow.

Dr Eric Verdaguer, Co-Founder & CEO, LogMeal

Abstract:

Hospitals and long-term care facilities need accurate information about what food is served, consumed and discarded. However, traditional monitoring methods are manual, inconsistent and difficult to scale. This talk presents a healthcare project developed with Agriculture and Agri-Food Canada in which LogMeal deployed an AI-powered monitoring system in hospitals and senior residences. The solution combines food images, depth-sensing data and before-and-after tray analysis to identify dishes and ingredients, estimate portion sizes, and calculate the amount of food consumed or wasted. The resulting data supports different professional teams: nutritionists can assess actual patient intake, operations teams can improve menu planning and procurement, and sustainability teams can monitor food waste and its environmental impact. The case demonstrates how multimodal AI can move beyond laboratory research and become an operational tool in complex healthcare environments, connecting computer vision, nutritional intelligence and real-world decision-making.

Bio: 

Eric Verdaguer is Co-Founder and CEO of LogMeal, an AIGecko Technologies company specializing in AI-powered food recognition and nutritional analysis. He leads the development and deployment of computer vision solutions that transform food images into structured information about dishes, ingredients, portion sizes, nutritional content and food waste. LogMeal’s technology is used in healthcare, senior care, digital health, professional sports and food-service environments.

Dr Jordina Torrents Barrena, Senior AI Manager and Principal AI Architect, AI Lab, HP Inc.

Jordina Torrents Barrena received her Computer Engineering BSc degree at Universitat Rovira i Virgili in 2014. She obtained her first MSc degree in Computer Engineering: Computer Security and Intelligent Systems (with honours) at Universitat Rovira i Virgili in 2016 and a second MSc degree in Computer Vision at Universitat Autònoma de Barcelona in 2016. She obtained her PhD degree from Universitat Pompeu Fabra on artificial intelligence, deep learning and medical imaging in December 2019. During the PhD, she performed secondments at King’s College London and Google UK. So far, she has published over 60 articles in scientific peer-reviewed journals and international conferences, 3 book chapters and has 6 patents. She was awarded the Google Women Techmakers 2017, Premi Dona TIC “Revelació” 2018, COE & Oracle Big Data and Artificial Intelligence Talent Awards 2019, McKinsey & Company “Next Generation Women Leaders” 2020, Ada Byron Award (Young) 2021, and Society of Women Engineering (SWE) Patent Award 2024, SWE Trailblazer Award 2025 and the Official College of Computer Engineering, La Nit Awards (Talent Jove) 2026, for her strong academic record, passion for increasing the involvement of women in computer science and demonstrated leadership. Jordina is currently the Senior AI Manager and Principal AI Architect at the AI Lab (HP Inc.). She is also an associate professor at Universitat Oberta de Catalunya and Universitat Ramon Llull.

Dr Lukas Fischer, Head of Applied Research / NXAI

Lukas Fischer is Head of Applied Research at NXAI, where he leads the development of efficient foundation models and industrial AI solutions. Previously, he served as Research Manager for Data Science at the Software Competence Center Hagenberg (SCCH), scientifically managing and coordinating the Data Science area with more than 50 researchers across multiple research groups. He holds a PhD in Medical Physics and conducted research at the Computational Imaging Research Lab (CIR) at the Medical University of Vienna, focusing on computer vision methods for medical image analysis and the quantification of trabecular bone microarchitecture. His research interests span foundation models, machine learning, deep learning, multimodal AI, computer vision, industrial AI, and trustworthy AI. Lukas has served as reviewer and program committee member for leading journals and conferences, including Nature Scientific Reports, IEEE TMI, IEEE JBHI, MICCAI, and MIDL. In addition, he has extensive experience coordinating and managing national and European research and innovation projects, bridging cutting-edge AI research with real-world industrial applications.

Dr Isabelle Siegrist, Founder & CEO at Sandborn, Investor

Isabelle Siegrist is the Founder and CEO of Sandborn, a Swiss company builder that helps entrepreneurs and corporates turn ideas into market-ready businesses, from idea validation and product development to market entry and growth financing. She has been building and investing in companies across Switzerland since 2014. Beyond Sandborn, Isabelle is an active startup investor and strategic partner at S2S Ventures, and serves as Strategic Advisor for Entrepreneurship at the ETH AI Center. As co-head of the Collective Gathering, she champions collaboration, diversity and female entrepreneurship across the Swiss startup ecosystem. She is also a coach at Innosuisse and regularly teaches entrepreneurship at universities. Isabelle was recognised as one of Switzerland's TOP 100 Businesswomen (2018-2020) and is a Forbes 30 Under 30 alumna (2019). She holds an M.A. in Business Innovation from the University of St. Gallen (HSG).

Dr Eric Verdaguer, Co-Founder & CEO, LogMeal

Eric Verdaguer is Co-Founder and CEO of LogMeal, an AIGecko Technologies company specializing in AI-powered food recognition and nutritional analysis. He leads the development and deployment of computer vision solutions that transform food images into structured information about dishes, ingredients, portion sizes, nutritional content and food waste. LogMeal’s technology is used in healthcare, senior care, digital health, professional sports and food-service environments.