View of the 3D Glyph Editor for creating and grouping glyphs.
View of the Binder interface for mapping data variables onto glyph elements.
Molecular dynamics data with the Conditional Box surrounding a region. Special glyph bindings may be chosen for the data in this region.
Data courtesy of Hai Ping Chen and Uzi Landman, Physics.
Researchers analyzing large datasets usually have neither the time nor the expertise to do their own graphics programming. Yet standard methods of visualizing data should often be tailored, or entirely new representations used, to achieve, in an effective and efficient manner, understanding of the data. Furthermore, the most creative analysis is an interactive process where the researcher refines her visualizations based on what she sees. We have developed programmerless, interactive methods to allow the researcher to construct glyphs, choose attributes, and then bind them to data. The binding patterns, ranges of change for an attribute, and numbers or types of attributes can be changed at will by the user. The results of changes are propagated automatically to the visual representation thus allowing a quick, iterative approach to finding the best depictions of the data. In addition we have designed these methods in a flexible, data-driven environment allowing user-defined filtering and manipulation and efficient instancing of thousands of glyphs and their attributes (e.g., pseudo-colored spheres representing positions and properties of each atom in a large collection). In their general form, these methods can build visual representations that are very rich and would subsume most current visualizations; they will also allow construction of untried representations or combinations of representations. We are now working on a major extension of these tools to allow effective study of correlations between several variables in spatially complex data. Our implementation of these methods, called GlyphMaker, is the first programmerless system for creating and editing glyphs and then binding them to data.
