Sergi Bermúdez I Badia

November 3


Currently, I am an Associate Professor (tenure) at the University of Madeira, where I teach for the Informatics and Interactive Media Design Masters; researcher of the NOVA Laboratory for Computer Science and Informatics and coordinator of its N-LINCS branch in Madeira, and president of the International Society for Virtual Rehabilitation. I received my Msc. in telecommunications engineering from the Universitat Politecnica de Catalunya (UPC) and a PhD from the Swiss Federal Institute of Technology Zürich (ETHZ).

I have pursued research at several institutes in Europe and the USA, including the  Laboratoire de Production Microtechnique at the EPFL (Lausanne), the Institute of Neuroinformatics at the ETHZ (Zurich), at the Institute of Audiovisual Studies at the Technology Department of the Universitat Pompeu Fabra (Barcelona),  where I was a Juan de la Cierva research fellow and head of the Robotic Systems Laboratory at the laboratory for Synthetic Perceptive, Emotive and Cognitive Systems (SPECS), and the Quality of Life Technologies and Entertainment technology centers of the Carnegie Mellon University (Pittsburgh).

My scientific goal is to investigate biological systems’ underlying neural mechanisms and exploit them using real-world artefacts, with particular emphasis on neuro-rehabilitation systems, interactive technologies, and robots.

Keynote title: Personalized Applications of VR to Stroke Rehabilitation and Fitness Training

Nowadays, it is widely accepted that games, and entertainment technologies in general, have very interesting features that, if used properly, can largely contribute to the effectiveness of treatments in different health domains. These games, also known as games-with-a-purpose, need to achieve a very difficult and interesting balance among science, health, engineering and entertainment. In this talk, I will present the approach we follow at the NeuroRehabLab, where we combine games, Human-Computer Interaction and clinical rehabilitation guidelines to develop interactive systems that are novel and effective tools for motor and cognitive rehabilitation, with special emphasis on stroke. I will discuss the effect of interface technology on motor-cognitive interference in task performance; a participatory design approach with health professionals to develop parameterized models for the training of Activities of Daily Living in a simulated environment; and how we automate the parameter selection process in these games by means of an adaptive approach. This strategy allows these systems to be used by patients with different cognitive and motor skills while still providing personalized training.

Abel Gomes

November 4


Abel Gomes is an Associate Professor in Computer Graphics and Games at the Department of Computer Science, University of Beira Interior, Portugal, and a senior researcher at INESC-ID (Graphics and Interaction Group), Lisbon, Portugal. He is also Associate Editor of Computers & Graphics (Elsevier) and International Journal of Computer Games Technology (Hindawi-Wiley).

His research areas of interest include Geometric Modeling, Computational Geometry, Computer Graphics, Computer Games, Computational Biology, and Molecular Graphics, and Augmented Reality in Medicine.  During his career, he has approached a number of scientific problems, namely convex hull algorithms, pathfinding algorithms, linearization of implicit curves, triangulation of implicit surfaces, triangulation of molecular surfaces, protein pocket detection methods, and quantum molecular dynamics. Currently, his main research topic is surface reconstruction using both classical and deep learning algorithms.

Keynote Title: Surface Reconstruction from Point Clouds: Past, Present, and Future

Reconstructing object/scene surfaces is an important, long-standing problem in computer vision and graphics research. Its applications range from computer-aided design, computer animation, robotics, virtual/augmented reality, or even medical engineering.  When dealing with this problem, we assume that the surface to be reconstructed from a point cloud is a 2D manifold embedded in the 3D Euclidean space. However, this problem is ill-posed because there are infinitely many solutions or surfaces for the same point cloud. Even worse is the fact that many point clouds may present noise, non-uniform point distribution, missing point chunks, outliers, and the like, as a result of some scanning issues during the data acquisition process.

Consequently, and considering both classical and deep learning algorithms, there is no reconstruction surface algorithm, capable of correctly reconstructing a surface from any 3D scanned point cloud. Surprisingly, or maybe not, a few classical methods perform even better than deep learning methods in terms of both robustness and generality. In a way, this talk aims to overview the state-of-the-art  in surface reconstruction, hoping to open some research avenues for the future.