[Back]

D. Kouril:
"Interactive Visualization of Dense and Multi-Scale Data for Science Outreach";
Supervisor, Reviewer: I. Viola, E. Gröller; Visual Computing and Human-Centered Technology, 2021; oral examination: 2021-04-07.

The progress in rendering scientific data has enabled researchers to visualize progressively more extensive and complex 3D models. Specifically, in biological visualization of the nano-scale, the challenge has slowly shifted from real-time rendering onto the issues related to understanding and exploring complex virtual environments built up by models consisting of, potentially, tens of millions of atoms. Considering that scientific visualization now increasingly serves to communicate the current knowledge to the general public, it is becoming more important to develop technology that allows people to interact with the visualizations.
This thesis focuses on the problem of navigating complex 3D models composed of large numbers of molecular instances packed densely in the three-dimensional space. Due to the large environment encompassing several magnitudes of scale, the traditional navigational paradigms applied in real-time computer graphics are becoming insuÿcient when applied to biological environments.
In the first part of the thesis, I analyze navigation challenges presented by such a use case and recognize several modes of navigation that can be employed when interfacing a complex 3D visualization system with the end-user. These navigational modes are implemented by exposing navigation-related parameters of the visualization to the end-user. Alternatively, control over these parameters is entrusted to the visualization system side and determined algorithmically. We discuss three such modes of navigation: augmentive, declarative, and automative. First, in augmentive navigation, the user is given fully manual control over all aspects of navigation, such as controlling the camera´s position and rotation, or the visibility of the individual model´s parts. Their manual exploration is, however, augmented by automatically deployed annotation to help make sense of the environment. In declarative navigation, the user interaction is simplified to declaring their target. The visualization system then takes over the low-level controls of the visualization, e.g., camera path animation and scene visibility transition, which are computed to navigate the user directly to their declared target. The third stage, automative navigation, relieves the user from even this responsibility and places the choice of what gets shown to an algorithmic solution. In this case, such automated fly-through can then be guided by a specific storyline.
In the second part of the thesis, I present specific methods addressing technical gaps and contributing to realizing the navigational stages presented in the first part. I start by
introducing an approach for textual labeling of multi-scale molecular models inspired by the level-of-detail concept. That way, a scenario of augmentive navigation is provided. Second, I propose a navigational method for traversing a dense molecular model with a hierarchical organization, implementing the declarative navigation concept. The presented method uses textual labels for browsing the three-dimensional model, essentially providing a way of traversing the hierarchical organization and exploring the spatial characteristics of the model. Finally, I propose a pipeline for producing automated tours of molecular models, demonstrating the automative navigation mode.

http://dx.doi.org/10.34726/hss.2021.91703