Selective Dynamic Manipulation of VisualizationsMotivationBarriersMore Barriers...Sample DataSDM componentsMore SDM componentsConstraintsFeedback TechniquesMore Feedback TechniquesHow Do We Use It?Favorite SentenceHow do we use it? (cont)Slide 14StrengthsWeaknessesContinuing EffortsExternalizing Abstract Mathematical ModelsThe ProblemThe SolutionOur ExampleThe Problems with Light BulbsMore Problems with Light BulbsIVA One: The Influence ExplorerThe Influence ExplorerInfluence ExplorerIVA 2: The Prosection MatrixProsection Matrix, contSlide 29Slide 30Selective Dynamic Manipulation of VisualizationsChuah, Roth, Mattis, KolojejchickMotivation•Need 3D techniques for interactive visualizations of multidimensional data. We want:–Selective: A high degree of user control–Dynamic: Interactions all occur in real time, with animation–Manipulation: Users can directly move and transform objects in the visualization•Author’s system is called SDMBarriers•Many data sets have too much information to be on screen at once•Much clutter and occlusion (hidden data) in dense sets of data•Difficult to give a sense of scale: some objects may be completely dwarfed by others (green objects in Fig 1)More Barriers...•Must be able to classify data into sets and save those classifications•Must be able to compare quantities which are not near each other (difficult to compare heights, for example, if they are at different distances from the user [Fig 3])•Authors believe SDM deals with these issuesSample Data•Crisis relief network–Supply centers are cylinders–routes between them are dark lines on the floor–shelters where supplies are needed are rectangular bars–heights of cylinders and bars indicate supplies available or needed–Arranged in a network, like Becker paperSDM components•Object centered selection–the selected set is made up of objects instead of a spatial area–can click on desired objects or use our old friend the constraint slider–when you create them, you can save and name themMore SDM components•Dynamic operations–The user uses a “physical” handle to manipulate the data (Fig. 4)–Attach a handle to an object, and push or pull on it: causes the object, or a set of objects, to grow, shrink or move–can control one or more parameters with single handleConstraints•Context persistence–SDM maintains a relation between the set being manipulated and the original set.•Set wide operations–if you can move or scale one object in a focus set, you can move or scale any.Feedback Techniques•SDM must clearly identify the selected set –so we know what objects will change if we take an action•SDM must maintain scene context –if we change something, a “shell” of the original value is left in its original place.More Feedback Techniques•Maintain Temporal Continuity–They use animation to allow the user to see what has happened without having to think too hard about it•Maintain relationship between selected set and environment–Keep a scale of the differences on screen, for example•Allow objects easily to be returned to their original positionsHow Do We Use It?•You can tell selected set apart by color or width•You can view occluded objects by –elevating them (Fig 9) - lose context–making all other objects invisible - lose context–making all other objects of height 0 (Fig 10) - lose context–make all other objects very thin (Fig 11) - still partially hidden–make other objects transparentFavorite Sentence•… the “physics” provided by SDM is not limited to real world manipulations; users can also elevate, compress, and perform other operations upon objects that wouldn’t be possible with actual physical models.How do we use it? (cont)•For different data sets, can use different scales–This is so that data sets with much greater or lesser values do not dominate•Can interactively make and visualize new classes of data–This is a lot better than having to update the entire database firstHow Do We Use It?•To solve the problem of comparing things at different distances, sets of data can be brought to the front and compared in two dimensions (Fig 13)Strengths•Enables a more precise, quantitative comparison between objects–preserves relationships between focus objects and rest of data–scaling is kept correct–distortions and occlusions of 3D are overcome•Also, it is pretty coolWeaknesses•Can only view limited part of the data set: the rest may be “in the distance” (possibly add rotation)•Can still get occlusion problem if focus set is dense•Does not address multimedia, UI, how to decide on representation?Continuing Efforts•Sage research project–SDM’s “physicalization” of the abstract space is combined with automated visualization tools, multimedia and UI stuff to create an entire systemExternalizing Abstract Mathematical ModelsTweedie, Spence, Dawkes, SuThe Problem•Mathematical models are important in many domains •They are often quite complex, not having an obvious physical visualization –an example of an obvious one would be a flow model might into a network or a pipe•How can we visualize them?The Solution•Interactive Visualization Artifacts (IVAs)–Instead of visualizing the raw data, we visualize precalculated data as 2 kinds of data•a description of the physical nature of an artifact, called parameters•a description of the results we can expect from an artifact, called performance criteria•We develop different IVAs to handle any given problem - we describe 2Our Example•The Light Bulb–design parameters: filament width, filament material–performance criteria: cost, brightness, lifetime•But there are problemsThe Problems with Light Bulbs•We need to create a light bulb given the performance data: but there is no way to get the parameters given the performance data (except trial and error - ugh!)•Changes in manufacturing mean that any set of parameters can only be guaranteed to be in a range of values - but not exact valuesMore Problems with Light Bulbs•Often, you also have to maximize some other objective, like manufacturing yield.IVA One: The Influence Explorer•We precalculate the data and display histograms based on it [Fig 6]•Each bulb design is represented once for each parameter and criterion: the design goes in the appropriate bin•The upper and lower limits on the sliders can represent the desired limits (Red passes all performance requirements,