Examples of our visualizations

We are developing a graphical software train simulator for virtualization of track conditions and the realization of test runs in a laboratory environment.

The simulator is a part of a research project the aim of which is to develop a functional sample of a rail vehicle detecting obstacles in the driving profile using an HW sensor system, sophisticated data processing architecture and with the help of artificial intelligence for their final identification and subsequent interpretation to the driver. The simlator is a key tool for the development and optimisation of the detection system.

For optimal utilization of supercomputers in numerical simulations of physical processes, domain decomposition methods are commonly used. These methods allow the division of large-scale models to achieve parallel scalability of a given computation or balancing of computational capacities. The video shows geometric domain decomposition applied to a jet engine. Domain decomposition is an important part of optimal algorithms capable of exploiting the potential of today's state-of-the-art HPC technologies.

In collaboration with medical doctors from the University Hospital Ostrava (FNO), we are developing tools for automatic segmentation and 3D reconstruction of tissue models from Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) images. We use machine learning and photorealistic rendering on HPC infrastructure of the resulting models to enable accurate and individualized analysis of medical data.

We develop novel methods and tools for acceleration of photorealistic rendering of massive scenes using Graphics Processing Units (GPUs). In our renderer, which is based on and compatible with popular Blender graphic suite, the scene size is not limited by the memory size of a single GPU because the scene data can be distributed among memories of multiple accelerators. By ading more GPUs, we do not only improve the rendering performance but also the maximum size of the renderd scene.

Given a sufficient number of GPU accelerators, the method is able to provide an interactive mode for rendering of massive scenes suitable for architectural visualizations, for instance.

The method is implemented at the memory management level, and thus it can be used with different path tracing-based rendering algorithms and their implementations.

Thanks to the HPC  infrastructure of our centre, we are able to perform production rendering of massive 3D scenes in a short time. We use the open-source Blender tool and the Cycles renderer, which we have modified for efficient use on the HPC cluster. In collaboration with the Blender Institute, we have helped to render several open-movies from their production in this way.

A customized volumetric rendering of four-dimensional image data from fluorescence light-sheet microscope enables to show the developing cell nuclei at high-resolution in a comprehensible way. It serves as a basis for further visual data augmentation, such as lineage tracking data, vector fields, or colour-coding of various cell quantitative parameters - a needed aid for developmental biology researchers.

Molecular Dynamics simulation of penetration of water molecules into bisphenol-based polymer at operating conditions (T=0-200 °C). Screenshot at the beginning of the 1ns-long run, with the water molecules being marked by the green colour and the middle region consisting of polymer. 

Security holograms are commonly used for the protection of goods, banknotes, and personal documents againstforgery. The fancy holograms are created by the combination of new optical effects and synthetic (computer-generated) or natural motives. Effects in final designed and fabricated holograms need to be easy to recognize by the human eye and hard to copy. The new effects are mostly based on fine and precise nanostructures. Virtual reality enables us to see the design of synthetic or natural motives and nanostructures before the prototype development.

The animation shows a simulation of a formic acid dimer at the temperature of 2 Kelvin using the classical Monte Carlo method. However, at such a low temperature, quantum phenomena that cannot be captured by classical simulation start to play an important role. It is therefore necessary to simulate such systems with methods that are developed to take these quantum phenomena into account. In such a simulation, hundreds of millions of values of the interaction energy of the forces acting between the particles need to be calculated in one run.