GILAB’s research lines include Videogames, Medical imaging, Scientific visualization, e-learning and Spatial and spatio-temporal data processing. Gilab responds to the demands of the industrial sector in different areas such as computer games and medicine.

• Global Illumination

Our research is focused on algorithms for the illumination of virtual environments. We aim to reach an equilibrium between realism and computational cost, considering the purpose of the generated images. Thus, while applications to interior design and animation require a high realism, videogames would rather need real-time computations. Techniques such as light-paths reuse allow for improving the performance of classic global illumination algorithms, based on ray tracing and radiosity. In the case of videogames, techniques involving GPU, simplification of algorithms, and non-physically-realistic rendering, are used to approach real-time computations.

• Videogames

During the last few years, the fast-increasing power of Graphics Processing Units (GPUs) has supplied videogames with the possibility of enormously increasing graphics realism. Our research is centred on the application of existing graphics techniques, or when necessary on the creation of new ones, for video games. This needs the integration of the techniques, once adapted or simplified to run in real-time, in the “game engine”, the software system that runs the game. Nowadays, game engines are not only used to make games but they are the basic tool in virtual reality systems, simulation, and virtual settings for movies and broadcasting.

Play The Urban Water Cycle or learn about Dry Stone Construction!

• Medical Imaging

GIlab works in collaboration with medical staff from the Institut de Diagnòstic per la Imatge, a prestigious imaging institute located in the main Catalan public hospitals, and with doctors from the Hospital Clínic de Barcelona. The goal of this multidisciplinary team is to research and develop useful solutions for the technical and clinical problems in the field of medical imaging and develop new tools to support diagnosis. Our research covers basic tools for diagnosis, computer-aided diagnosis, and diffusion tensor imaging.

• Machine Learning

We focus on machine learning with data uncertainty and the use of probabilistic graphical models (PGMs). These PGMs offer interpretability and can handle partially missing information. Learning methods must be specifically designed following strategies such as Expectation-Maximization, and when exact inference is unfeasible, reasoning should be done approximately using, for example, variational or Monte Carlo methods.

We have applied these techniques in healthcare (to predict problems and select embryos), in community modelling (to understand relationships and collaborations), and in environmental research (to address issues such as beach pollution and fish management).

• Scientific Visualization

Scientific visualization describes the field of computer science which deals with the study and definition of algorithms and data structures for the visualization of scientific data. Visualization has become one of the most important ways of exploring data and it is applied in many scientific fields and application areas, such as material sciences, fluid dynamics, environmental sciences or medicine. In our group we investigate visualization and simplification techniques for exploring 3D medical data.

• Viewpoint Selection and Shape Recognition

Viewpoint selection is an emerging area in computer graphics with applications in fields such as scene exploration, image-based modelling, and volume visualization. We investigate viewpoint quality measures to select the best views and to explore a scene. We also study the information and saliency associated with the different parts of an object. Some of those measures are used to simulate the illumination of a scene. In the object recognition field, we are working on the definition of shape descriptors based on uniform distributions of lines and viewpoint techniques. These descriptors permit us to calculate the similarity between objects of large databases.

• Computational Aesthetics

Computational aesthetics joins together different fields such as computer science, philosophy, psychology, and the arts.  In particular, we investigate informational aesthetic measures to quantify order, complexity and information in images, paintings, and sculptures. The used techniques are mainly based on information theory, Kolmogorov complexity, and viewpoint selection quality measures.

• E-Learning

E-learning is a very extensive research field. Our group focuses on developing specialized web-based tools for the automatic correction of complex open-answer activities. We develop correctors for math, programming, database design, diagrams or graph exercises within this context. These correctors are integrated into the ACME e-learning platform. Currently, this platform has more than 4.000 users and is applied in more than 40 different subjects. Our investigation is also focused on automatic assessment.

• Spatial and Spatio-Temporal Data Processing

Research in this area is focused on spatial and spatiotemporal data mining with the aim of facilitating decision-making in fields such as facility location, competitive-collaborative marketing, vehicle traffic management, urban planning, touristic development, medicine, epidemiology, and the study of social interactions. We develop algorithms and data structures to extract meaningful non-explicit information from spatial databases. As experts in the area of Computational Geometry and the use of the GPU, and facing the challenge of Big Data, we seek efficient solutions which usually are, totally or partially, obtained in parallel.