Conveying Shape: Lines
From Visualization Sp06
Lecture on Mar 14, 2006
Readings
- Automatic illustration of 3D geometric models: Lines. Dooley and Cohen. (acm)
- Line Direction Matters. Girshick et al. (pdf)
- Suggestive contours. DeCarlo et al. (html)
Optional Readings
- Illustrating smooth surfaces. Hertzmann and Zorin. (html)
- Automatic illustration of 3D geometric models: Surfaces. Dooley and Cohen. (pdf)
- Speed of Perception as a Function of Mode of Representation. Ryan and Schwartz. (jstor)
- Assessing the Effect of Non-Photerealistic Rendered images in CAD. Schumann et al. (html)
Contents |
Bryan - Mar 16, 2006 07:22:17 pm
The suggestive conours paper was my favorite at SIGGRAPH '03... I think it was the closest thing to what a human doing line drawings would do that I've ever seen. It reminds me also of a neat software package called SketchUp. It's an amazingly easy to use 3D modeling package that explicitly includes "sketchy" views of models that are designed to encourage revision over traditional architectural renderings. Unfortunately it's not free, but I think there's a demo version you can download and play with.
Cynbot - Mar 18, 2006 09:39:58 pm
I was so impressed with "Illustrating smooth surfaces ". I wonder if these direcional line drawing algorithms could somehow be encoded as descriptors of shapes for their specification in searching databases.
So if you were searching google images for a photo of a car taken at a given angle or under certain lighing conditions you could just have a knob (like the photoshop drop shadow angle specifier) setting maybe to a first order a gross curvature or lighting angle.
Or in 3D, one could image adding to a sketching search interface a vector filed of these hatches.
Gwyu - Mar 19, 2006 09:17:20 pm
Though this point isn't exactly visualization-related, I think the paper on non-photorealistic rendering in CAD is interesting as it seems to support HCI design practice, where lo-fi prototypes are preferred early on not only because they saved time by not wasting effort implementing things that will be changed later, but also because they encourage test users to view the prototype as something that is still in the formative stages. The makers of Denim found similar practice among web designers too, though there was also a balancing act between conveying a sense of malleability and a sense of professionalism.
Brien - Mar 20, 2006 12:35:07 am
I think this topic – lines and how to summarize complex objects – is a really cool mix of art and geometry. I guess this whole non photo realistic section has had a similar flavor, which I like.
I remember a family of surfaces called minimal surfaces (GRAPE visualizations) that have zero mean curvature. There are some pretty weird surfaces in this family (e.g. Enneper's), but a plane is also minimal. If I remember correctly, principal directions are undefined when there is zero mean curvature. So principal curvature techniques for showing the surface would break down when the surface becomes near flat? I think the DeCarlo et al paper mentions this case. Further clarification would be appreciated.
Maneesh - Mar 20, 2006 05:36:47 pm
Brien - Directions of principle curvature only exist at points for which the Gaussian curvature is not 0. For most points on smooth closed surfaces the Gaussian curvature is non-zero and therefore principle directions exist. Note that the mean curvature and Gaussian curvature are slightly different.
Neat visualizations of minimal suraces!
Gwyu - I too think that the tension between malleability and professionalism that is conveyed by visual design is interesting. There has been a bit of follow-on work to the CHI paper, but I haven't seen a very thorough treatment of the issue.
Sharena - Mar 20, 2006 06:43:59 pm
Thinking about the Architecture drawings in the lecture and the effects of the different drawings styles seems very interesting to me. I've done paper prototypes and I've heard about how the low fidelity prototypes are good for making changes, etc so I understand the results of architecture survey in that respect. But I also though it was interesting to see how useful the CAD diagram was over the shaded image in the final presentation - it shows where the lines are more useful than a fuller picture. Its clear how lines are good enough for prototyping, but what I found interesting in this topic was how they can also be useful as final presentations - like the cartoon drawings shown in the beginning as well. I had not thought about lines by themselves that much as a tool for making things easier to understand.
Raymond - Mar 21, 2006 12:01:37 pm
I remember reading a SIGGRAPH paper last semester for ideas on my CS184 final project- and I came across one that employs suggestive contours and uses it as artistic strokes. For shading, it also uses some type of stroke and thus, the entire image looks handdrawn, when in fact it is a complete 3D model! It was a cool look in NPR... sadly, I forget the name and author of the paper.
Mattkam - Mar 21, 2006 11:51:10 pm
Of course, one caveat with using lines and lo-fi prototypes is that the test user must be able to visualize how the actual object looks like. This can be difficult if the test user does not have adequate familiarity with the object being referred to. For instance, in our fieldwork with rural children in India, we quickly learned that lo-fi prototyping wasn't effective because they found it hard to imagine what the eventual hi-fi software prototype would look like.
