Conveying Structure

From CS 294-10 Visualization Sp10

Revision as of 00:18, 15 May 2010 by Shimul (Talk | contribs)
(diff) ← Older revision | Current revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Lecture on March 29, 2010




  • Smart Visibility in Visualization. Viola and Gröller (html)
  • Using Deformation for Volumetric Browsing. McGuffin et al. (pdf)
  • Interactive Image-Based Exploded View Diagrams. Li et al. (html)

Optional Readings

  • Interactive Cutaway Illustrations of Complex 3D Models. Li et al. (html)
  • Non-Invasive Interactive Visualization of Dynamic Architectural Environments. Niederauer et al. (html)

Boaz Avital - Mar 31, 2010 07:41:06 pm

What strikes me most about these 3d visualization techniques, besides how cool they can look, is that they have a significantly higher barrier of entry than the techniques we've learned up till now. Most of what we've seen is fairly simple for the person creating the visualization, allowing them to simply plug in their data or a modified version of their data along with some specifications or logic in how to display it. For these, it seems you would need a dedicated graphic artist and/or dozens of hours designing and specifying the 3d models in order to produce something useful. The end result of though is something capable of conveying quite a lot of information in a form that's of course easy for the viewer to understand. What work, if any, is being done to make this avenue of visualization more accessible?

Jon Barron - Apr 02, 2010 02:52:36 pm

Viola and Gröller:

Where's the beef? I don't really see an actionable algorithm here, just the idea that "important" things should be more opaque, and the unimportant things in front of them should be more transparent. Still, I suppose it's primarily a survey paper, and I do like most of the visualizations they present


I found it very helpful to watch the video demo of this project:

I really like parts of this paper, but watching the video, only the hinge tool seems at all useful, probably because it has the most direct and intuitive physical analogy. Everything else seems pretty unnatural, and not more useful than the hinge tool.

The interface widgets also seem far too complex (though again, the hinge tool isn't bad). Not that they're poorly designed, but it seems that they *must* be complex, because the operations they're proposing just have so many degrees of freedom. It would be cool if they took the hinge widget and just built a physical interface into it. It could be pretty simple, and the benefit to the user compared to this awkward widget interface would be huge.

Li et al:

This is a very neat idea, though it's very much a boutique problem. It seems to me that if you're in the position to author an interactive diagram like this, you probably have access to the full 3D model of the object you're diagramming, and so it may just make more sense to impose these inter-object sliding-hinge relationships in 3D, and just present an interactive 3D model. This would circumvent all of the depth-ordering problems.

Zev Winkelman - Apr 02, 2010 03:33:24 pm

Viola and Groller:

Smart visibility "exploits the expertise of the human observer" (2) - does that mean that the generalizability of smart visibility techniques will require at least one smart human per domain ?

Using "importance" to encode which features are most interesting, and "importance compositing" to determine the rendering, of course, leaves the subjective assessment of importance up to the expert human (3.1) . This might work in fields in which there are consensus about what is important, but might not when this question is contentious.

I thought the paper was a good overview of various technical illustration methods. I had not read about, or at least no known the technical terms for these methods until I read this article.

Niederauer et al:

I think the coolest part about this is that it works "noninvasively" at a layer that the application isn't quite aware of. After seeing the demo in class I think the augmented spectator mode would be really cool. Another interesting challenge might be to remove the need for user input regarding the number of stories.

Also might have to look into Chromium some day.

Li et al:

I thought the demo in class of this system was really cool. I am complete agreement that allowing interaction dramatically increases the effectiveness of the display. The authoring pipeline was also very impressive.

McGuffin et al:

Animations have been referred to several times, but I appreciated the reference the authors give for work that explicitly establishes the benefits to users of having to spend less time assimilating new states (3 -Bartram 1997, Grossman et al 2001, Robertson et al 1991, Woods 1984).

I'd seen something like this at the Immersive Visualization Lab at UCSD a year ago, but had not explicitly thought of the benefits of deformations over cutting in that they offer an alternative that doesn't necessarily sacrifice any context in order to look inside.

Jonathan Yen - Apr 03, 2010 10:58:55 pm

Smart Visualization: Along with what Jon was saying, it doesn't look like there's anything that computers can do to produce these kinds of visualizations--for the most part, it looks like it all just comes down to an artist's perspective of what occurs at the inner levels of presentation. Overall though, I think this paper provides a decent overview of the different ways in which we can produce more expressive visualizations: volume cutting, exploded views, cut-away views, section views, ghosted views. I think the document browsing application depicted in the paper and other applications of this smart visualization technique are interesting and worth further study.

Volumetric Browsing: The different tools that are presented in this paper are interesting. I can see how one tool might be more useful than another, depending on how one is trying to use the deformation to get a different perspective of the image. I wonder about the ease of use for each of these tools--is it simple to use these tools, particularly given that one would be operating in a 3D environment? My experience with a lot of 3D applications is that it can be tricky to navigate through the space and to process the data.

Image-Based Exploded View Diagrams: My initial impression after looking at this paper is that it looks like it would make for a very helpful educational tool. The interactivity of these diagrams can be useful for training, especially for work which involves many complex parts, such as being a mechanic or a repair technician. I wonder if further work could be done for 3D rotation, so that people can get a view from a different angle for better clarity and perspective.

Stephen Chu - Apr 04, 2010 01:31:44 pm

Viola, Groller: This paper provides useful rules to follow when creating visuals using smart visibility techniques. Improving these techniques while allowing interactivity will greatly help the teaching of surgical-like operations. Diagrams in biology and medical textbooks can show great detail, but their lack of interactivity is a huge hindrance. By giving students an interactive visual of a biological system, the students will be much more knowledgeable and familiar with the system before actual hands-on work. E.g. Even though we were provided a detailed cut-away view of a rat in one biology class, many students were confused when looking at the actual rat dissection. An interactive visual tool would have helped in familiarizing students before the lab.

Arpad Kovacs - Apr 05, 2010 01:51:19 pm

Viola, Groller: This survey paper provides a broad overview of previous work in smart-visibility techniques such as cut-away, section, ghosted, exploded views, and deformations, as well as high-level algorithms/guidelines for creating effective transparency and assembly instructions. We covered many of these ideas in lecture already, however one approach I have not seen before is Bruckner et al's illustrative context-preserving volume rendering technique. As used in Volumeshop, this mapping of transparency to the strength of specular highlights, along with the ability to specify depth parameters for "fuzzy clipping", appears to be a novel way to enable interactive exploration of the insides of ghosted models while still providing the user with depth-cues. It would be interesting to see whether these enhancements can make ghosted models competitive with cutaways in practice, or if the ghosting technique's shortcomings are still too great to overcome.

McGuffin et al: I thought that this was a novel approach to representing volumetric data by distorting voxels in an octree datastructure. I agree that the hinge and leafer tools are quite effective, since there are clear physical analogies for these actions, and they enable detailed examination of individual layers. However, for the box spreader and sphere expander, I think that simply removing voxels would have been a better solution than the demonstrated technique of pushing voxels aside, which does not reveal much information about the outer layers, but may occlude the areas of interest. Since this is a "smart visualization" technique, we can assume that the viewer has expertise in the domain, and can still mentally complete the object from partial information.

Li et al: I thought that this interactive visualization tool was very useful and has many practical applications; it would be especially valuable for creating interactive diagrams of older exploded-diagrams, which were not designed using modern 3D CAD tools. Annotations and text-based search for parts was a very neat addition, since this seems to be a direct implementation of Schneiderman's Visual Information-Seeking Mantra of overview first, filter, then details-on-demand. For future work, I wonder if this system could be automated even further, for example by infering movement and position constraints from preexisting guidelines on the diagram. Additionally, it seems that without a 3D model, there is no way around manually specifying the occlusion relationships; however the system could suggest locations for occlusion boundaries by comparing the sizes of spatially adjacent, overlapping parts (eg: when a smaller cylinder is covered by one with a larger diameter).

Akshay Kannan - Apr 15, 2010 07:12:21 pm

The Viola/Groller paper had some very interesting examples of 3d cutouts. I also find the idea of "Smart Visibility Techniques" interesting, by which less important elements are occluded and symmetry and human cognition are exploited by showing only half of a symmetrical image, and allowing the observer to infer the entire structure. As for the decision between using cut-away and ghosted views, one of the primary factors seem to be the depth vs breadth of the 3d structure. For skinny "deep" structures, consisting of a large depth of layers, with limited information on each depth, a cut-away approach seems to be superior, since multiple layers can easily be portrayed without any distortion, and since only limited detail needs to be shown on each layer, the illustration remains clear. As for wider structures, consisting of only one of two layers, with a lot of detail in each layer (for example, the circulatory system underneath the visible human skin), a ghosting approach tends to work well.

Prahalika Reddy - May 12, 2010 07:41:06 am

I find the topic of conveying structure very intriguing. It's something that I've felt is very difficult to do effectively. The examples we looked through in class were very interesting and had some very good points about each technique for conveying structure.

Personally, I like transparency and cutaways best for conveying structure. I think it shows the inside of structures without disassembling the structure, which is very important. I'm not a big fan of exploded views because of that reason. Exploded views make it hard for the viewer to picture what the structure looks like as a whole.

Priyanka Reddy - May 14, 2010 06:20:32 am

I thought the idea of taking a static image and turning it into a dynamic, exploded view was really interesting. However, doing so does require lots of user interaction, and the final result isn't perfect. In fact, I can't image there would be many pictures that work as well as the one we saw in class. Although I think it's a pretty neat method of creating visualizations, I think its scope of usage is really small.

The leader tool was cool and interesting to play with from a technology standpoint. However, aside from that, I'm curious about who this would actually benefit. I think it's too much detail for a layman, but not enough for doctors and premeds.

Shimul Sachdeva - May 14, 2010 06:48:19 pm

The exploded views seem to be a very useful way to visualize 3D models. Interactivity with these models makes for a great visualization tool. I wonder how hard it is to create such models. Are there software programs like Tableau for 3D visualizations as well? Is this a category shared by CADs?

Layering is also a good way to highlight important information in a dense visualization. I found the concept of cut-away views in the "Smart Visibility in Visualization" paper interesting. The paper on "Interactive Cutaway Illustrations of Complex 3D Models" has some great examples for medical purposes. The authoring tool looks pretty awesome - can we download the software?

[add comment]
Personal tools