XML3D (3D Internet)

„XML3D specification: Developed and published specification, implementations, tutorials and examples. Started W3C Community Group.“
„Simple and Robust Iteractive Importance Sampling of Virtual Point Lights“
We have seen two major trends in IT over the last two decades: The growth of the Internet as the dominant communication network and the emergence of high-performance hardware for 3D graphics. Surprisingly, given the importance of both events, there is hardly any 3D graphics content on the Internet today. Integrating interactive 3D graphics into common browsers would bring these two major trends together and open up an entire new platform for the development of applications with 3D capabilities.
Intel VCI, Saarland University, and the DFKI jointly develop a base platform for the 3D internet, core of which is a new declarative scene description language XML3D. XML3D is an extension to HTML5 which enables processing and displaying of bleeding-edge interactive 3D graphics in the browser. The integration with HTML as well as other ubiquitous W3C standards such as DOM, CSS, and others, allows millions of web programmers to directly apply their existing knowledge and easily create interactive high-quality 3D content and applications.

- This chess game demonstrates the interaction capabilities of XML3D. In this examples, a chess piece can be moved by first selecting it with a mouse click and then by clicking on the target at the chessboard. While moving the mouse pointer over the chessboard, the currently pointed element is displayd in the topmost box.
In this project we are building upon the XML3D platform, improving and extending it in two major areas:
- advanced ray tracing and
- a shader-based data flow computing capabilities.
In the area of advanced ray tracing we aim at developing a flexible rendering system that supports realistic light simulation and complex shading, while retaining both interactive performance and high quality. Using an anytime approach we are able to interrupt the computation so that the user always experiences full interactivity but we keep refining the results to also produce the highest quality possible. In this context, we focus on studying and improving progressive virtual point light source techniques, research of advanced geometric primitives, and parallel traversal and build algorithms. We also put emphasis on evaluation and optimization of these algorithms with respect to key features of modern multi- and many-core architectures, such as memory coherence and bandwidth requirements, SIMD utilization, or mapping the algorithms to massive parallelism.

Besides the improvements in core rendering, we also extend the platform with shader-based data flow computing capabilities such as animation and geometry processing like morphing and skinning or digital media (post-)processing. We aim at a general extension to XML3D scene graph with a dataflow-based computational model that enables flexible and fully programmable description of the processing pipelines. The representation is designed such that it can be adaptively compiled into a form optimized for multi- and many-core hardware, specifically addressing efficient scheduling and memory management when passing data through this pipeline. We also require these entire graphs are made accessible within the DOM so that any browser application can create, modify, and delete them, thus integrating well with other native element types within a XML/HTML document.
Project team
Principal Investigator
Prof. Dr.-Ing. Philipp Slusallek
Researchers
Sergiy Byelozyorov
Tomas Davidovic
Dipl.-Inf. Georg Demme
Stefan John
Javor Kalojanov
Felix Klein
Lukas Marsalek, M.Sc.
Dr. Vincent Pegoraro
Dr. Kristian Sons
