Speaker
Description
The talk is focused on scientific visualization. High-fidelity scientific simulations utilize HPC infrastructures and generate terabytes to petabytes of data. To properly analyze and explore such data fast post-processing and visualization are fundamental.
Two approaches, which are developed to provide a workflow for high-quality cinematic visualization of large scientific data sets, will be presented. Both utilize IT4Innovations infrastructure to ensure data pre-processing and rendering. Blender software is used to create final visualizations. In this way, numerous features such as lights, materials, cameras, and others can be conveniently set up. Moreover, we have developed a modified version of Blender path-tracing-based renderer, which allows the distribution of the computation among cluster resources and thus provides high-quality as well as fast rendering.
The first approach is focused on visualization using a path-tracing volume rendering technique. A scalable workflow for processing and visualization of unstructured data will be described. This approach uses our internal MESIO library for data set loading and processing. The second approach integrates Vistle into Blender. Vistle is an open-source tool for scientific data reading and processing. Blender serves as a user interface with further rendering capability while Vistle runs in the background and can utilize HPC resources for data processing before the visualization.
