Interactive Visualization Techniques for Conveying 3D Structure
From CS294-10 Visualization Fa08
Lecture on Nov 12, 2008
Readings
- Interactive Image-Based Exploded View Diagrams. Li et al. (html)
- Interactive Cutaway Illustrations of Complex 3D Models. Li et al. (html)
- Automated Generation of Interactive 3D Exploded View Diagrams. Li et al. (html)
Seth Horrigan - Nov 12, 2008 01:46:42 am
I have seen a few of these projects as well as some related projects in previous presentations that Maneesh has given. These papers distill key concepts in viewing "inside" objects, then provide automatic computer-aided techniques for employing those best practices. In the description of exploded views, the idea of canonical explosion directions caught my eye. It occurred to me that almost all exploded views I have seen are along only one or two canonical explosion directions, and this improves the perception of how the object fits together. I can imagine objects where there might be many canonical explosion directions (e.g. imagine one of these with complex components distributed along the axis of each spike: http://www.joissu.com/prodinfo.asp?number=29-113), but I also wonder if there are ways to convey the assembly clearly without using the canonical directions?
I suppose the guidelines research from Maneesh referenced in that paper might allow that, which could allow more components to be visualized in the same space (since they can expand anywhere in the x-y plane, just like an actual explosion, instead of along specific axes). The system described does seem to deviate somewhat from the canonical axes though, especially when splitting "sub-assemblies". I do not mean to imply that the current techniques for exploded views are lacking; rather, the paper got me thinking about other possible adaptations of exploded views.
Edit: I just noticed the "Arbitrary explosion paths" section in the '04 paper (I think we read it before) that mentions the exact idea I posited above, "To achieve a more compact exploded view layout, illustrators sometimes arrange parts using non-linear explosion paths that are often indicated with guidelines."
Scott Murray - Nov 12, 2008 02:49:36 pm
The work presented in "Interactive Cutaway Illustrations of Complex 3D Models" is groundbreaking, but the static format of a research paper doesn't do it justice. It's necessary to watch the demonstration video to fully grasp the dynamic nature of the viewing tool and to understand how watching component parts assemble (and disassemble) in time contributes to the user's understanding of product assembly.
That said, I'm left with the same question that I have after reading most of these papers, which is: "What does the user interface look like?" The authors have unarguably built a new, helpful visualization tool, but we have to take them at their word that these 3D models can be created with "little user effort" of "just a few seconds per part." The demonstration video skips over the authoring step and focuses on the viewing interface. But to my mind, the authoring component is just as important, for the tool's usefulness hinges on its usability. A poorly implemented authoring interface would inhibit the tool's use, while an intuitive one would only bolster the author's argument in support of its usefulness. Why not show us the interface in addition to the end visualizations? Maybe that's what we'll see in the lecture today.
Ketrina Yim - Nov 12, 2008 07:14:15 pm
@Seth: For illustrating a spiky bounce ball with a complex structure inside every spike, I'd think the best approach would be a cutaway to show context and an exploded view of one of the spikes for detail. Exploding the whole thing would yield an outwardly radiating shower of parts that could be hard to read (and guidelines would only make it worse). Of course, this would only work if every spike had the same components, or if the inner components came in only a few variations.
After seeing the 3D cutaways and exploded views, I can't help but ask a one question from a technical standpoint. What file types do these visualization authoring tools take as model data? Due to the illustration-like style of shading, I can't tell if the models are composed of polygons, NURBs, or splines and patches. I'd also ask whether one would use modeling software (like Maya) or CAD to create the input models, but I figured the answer lay in the file types accepted.
Simon Tan - Nov 18, 2008 12:08:57 am
Li's application for generating interactive image-based exploded view diagrams was quite impressive. It went beyond the basic work of separating an image into sprites for sliding; it also took care of all the odd occlusion problems that result when you animate the movement of one sprite over another. The end result was so realistic that at times I couldn't believe I was just watching two-dimensional sprites sliding around. I'm sure the authoring interface could be used in some other context to produce three-dimensional effects for two-dimensional images.
I also wanted to applaud the use of acceleration in the exploded-view animations (for the 'explosions') - it was definitely a nice touch to make the animation more physically realistic, naturally guiding the viewer to watch along a primary axis.
Michael So - Nov 18, 2008 10:22:49 pm
The 3D interactive exploded view diagrams are interesting. The interactivity and filtering techniques I can imagine to be practical because it is the sort of thing a person would do if they had the real object of interest in their hands. Or actually, having an exploded view diagram can be more practical to use to understand an object than just playing around with the real object in your hands. In the exploded view diagram, everything is organized and you get an overview of everything and idea of spatial relationships between parts. By making it interactive, it helps reveal and either confirm or correct how one thinks what the spatial relationships are and how parts fit and relate to one another. Probably having both an interactive 3D interactive exploded view diagram and the actual object in one's hands would be the best way to learn the structure of the object. Because with the diagram, you get the animation and the overview and the ability to filter to emphasize only certain parts. But with the object at hand, you get the physical information on it (like how it feels and how it compares to the size of your hand for example). Plus you can open up the real object and confirm what the diagram dictates to you.
One question I have about the system is if there is any way to alter the speed of the "explosion" animation. Because at some points, it might go "too fast" and it might be helpful for understanding if there was a way to "slow down" certain explosions. Especially if in the future if someone plans to include certain complex interlocking relationships (i.e. a screw being rotated in order to be removed), it might be useful to slow things down to notice these kind of relationships between parts.
Dmason - Nov 19, 2008 01:24:36 pm
I found the lecture very promising, and it made me wonder where the idea and need for such work came from. Thus I sleuthed a little bit and found the full repository on this fellow's papers: http://www.cs.washington.edu/homes/wilmotli/ This is the same fellow, by the way, who developed the automatic layout generation.
We discussed in class how these tools can handle a variety of situations -- given exploded-view diagrams and rendering them interactive, or providing a framework to develop your own exploded-view diagrams. It all makes me wonder what the commercial need for these diagrams is... so I also did a little googling and came across some interesting applications that weren't explicitly mentioned. These kinds of diagrams can also be used to explore graphical data, such as pie charts with subslices coming out of each one, as well as geographic data. I would be very interested to see a more thorough inquiry into these applications.
Calvin Ardi - Nov 19, 2008 02:55:42 pm
Wilmot Li's lecture was very well done, especially with some of the demonstrations presented after the concept was covered. I do agree with the comment presented in class regarding the rather slightly grotesque imagery when "peeling" away features of the human head. In that respective paper, it was interesting to note that after several user studies, it was found that having an animation between states (e.g., peeling away the skin) was very helpful in keeping track of changes between states. Intuitively, it does make sense; I'd find it a bit hard (and perhaps surprising) to have an image jump back and forth between a model with the skin and one without (revealing it's underlying structure). For those who are a bit impatient, does the tool support control over the animation speed?
I especially liked the tree/graph structure of the explosion graph and the overall approach of generating these interactive 3D exploded view diagrams. I don't recall if this was covered or brought up during the lecture, but it would be neat to have the tool support rotation/zooming of the 3D model.
(Also, slide 61 of the lecture is broken: a lot of the text overlap with each other).
Yuta Morimoto
In presentation, there are many interesting things, especially making directed acyclic graph of a structure of 3D object. Initially, I think the application need to use some hierarchical approach to decompose and decompose the object. During presentation, we saw how the structure and relationship made and I thought some questions on it. If we can have a DAG of some product, we apply the structure to make new one. I have a no idea about I, but it might be interesting.
Chris - Nov 19, 2008 04:42:56 pm
Regarding the Interactive Cutaway Illustrations of Complex 3D Models by Li et al, I would be interested in seeing if there were a reliable way to perform this type of visualization on real-world data sets. By real-world data sets, I mean volumetric data coming from, for example, medical scans. In particular, considering how effective this cutaway view is for the synthetic anatomical model, I would imagine it to be useful as well for pre-operative exploration, say, for surgery. One of the principal issues that would come up, which would need to be solved, would be automatically segmenting the scan data into individual components to be visualized.
HeatherDolan - Nov 20, 2008 012:12:07 am
I really enjoyed this lecture and the associated papers. I would like to see some image-based diagrams created from other sets of files (photos). It looked easy in the demo, but I am left wondering how much effort went into creating the photographs themselves. That's not to say that the initial work to take quality photographs isn't worth it. The results of carefully taken images and the tools described may still be efficient compare to an entirely manual (hand drawn/modeled) process. However, I can imagine different users for a tool of this type who may or may not be inclined to take the right kind of photographs to produce the quality results we saw in the demo. That being said, it's a skill that can be developed.
I liked the interactive viewing as well. The animation looked and "felt" nice and somehow natural as did the selective expansion/collapsing. Several of the future work topics sound like nice enhancements (Emphasis, Semantic Zooming in particular).
Witton Chou
I would probably have had a much easier time in high school biology (the last time I even touched biology) with these tools. The interactive cutaway illustractions of complex 3d models really caught my attention. Three dimensional models are great, but being able to interactively reveal interior constructions through simple rigging is phenomenal. If these tools were available when I had to learn the human anatomy, these would be much superior to the pixellated photocopies we were provided where we couldn't tell what we were looking at half the time.
What is very impressive about these utilities is the ability to recognize construction patterns and selectively reveal and peel back layers that obstruct the view of other important features. I really enjoyed Wilmot's talk about these utilities and hope to see more of these tools be used for both educational and commercial media.
Sarah Van Wart - Nov 24, 2008 11:17:16 pm
Wilmot's lecture gave the class a more comprehensive look into the different interactive layering visualizations that we have been reading about over the past week. It was nice to have supplemental by live demonstrations of how interactive cutaways are implemented in various dynamic software packages. Since my final project with Razvan is an animation project, I paid particular attention to the way in which animation was used to move source parts from their original orientation to the foreground as the user drilled down into the data. The animation really allows the user to see how parts fit in a way that simply having the new view of a part "materialize" in the foreground. I used to think of animation in graphical interactions as little more than eye candy, but I am coming to realize from a number of examples this semester that it's a very powerful technique in preserving continuity and orientation.
