View source

Lecture on Dec 1, 2008

[http://vis.berkeley.edu/courses/cs294-10-fa08/WWW/lectures-WWW/shape-shading''Slides'']

===Readings===
*Perceiving Shape from Shading. Ramachandran ([http://psy.ucsd.edu/chip/pdf/Shape_From_Shading_Sci_Am.pdf pdf])
*Using Non-Photorealistic Rendering to Communicate Shape. Gooch and Gooch. ([http://webhome.cs.uvic.ca/~bgooch/Publications/NPRcourse_sec8_gooch.pdf pdf])
*Conveying Shape and Features with Image-Based Relighting. Akers et al. ([http://graphics.stanford.edu/papers/ib-relighting/ html])
''Optional Readings''
*The lit sphere: a model for capturing NPR shading from art. Sloan et al. ([http://www.cs.utah.edu/npr/papers/LitSphere_HTML/ html])
*Automatic lighting design using a perceptual quality metric. Shacked and Lischinski. ([http://www.cs.huji.ac.il/labs/cglab/papers/ldesign/ web])
*Maximum entropy light source placement. Gumhold. ([http://ieeexplore.ieee.org/xpl/abs_free.jsp?arNumber=1183785 ieee])
*Light Collages. Lee et al. ([http://www.cs.umd.edu/gvil/projects/light_collages.shtml html])

== Nivay Anandarajah - Dec 01, 2008 02:11:57 pm ==

The Perceiving Shape from Shading article illuminated how subtle changes in how we present our shaded items can entirely change the perception of their solidity.

What I want to know is common guidelines for using these subtle changes in creating an effective GUI. The more people get access to cracked versions of photoshop, the more every interface you see has a belligerent use of drop shadows, embossing, and gradients. It's tough to find a website that doesn't have some form of a gradient on it now a days. This might just be the current trendy practice - a naive approach at "pleasing aesthetics." But I assume there must be some added usability to the visualization to create the illusion of depth. By using these principles of single light source shading, designers can create shading with a purpose. For example, the direction of the drop shadow on a button totally changes its perception of use. If the shadow is receding to the bottom right, the button is perceived to be floating. If the drop shadow is receding to the top left, the button appears to be embedded in the surface.

Do people want embossed buttons to mimic a tangible user interface? Does adding a small gradient to every tool bar help add a sense of layering and differentiation? Does adding a drop show receding to the top left give the user a hint that it is clickable? This perceived shape from shading applied to usability would be an interesting topic.

== Scott Murray - Dec 01, 2008 04:52:51 pm ==

Very interesting questions, Nivay.  I totally agree that gradients are far overused in most UIs.  They may demo well (as in, "that looks cool"), but are not very practical for most purposes.

I thought that today's discussion was very interesting and highlighted a number of interesting perceptual issues, but it's not clear to me yet how the issues of shading can translate into resolving general data visualization problems, except to say that "it's complicated" and, like good lighting design, may take some experimentation.  I'm trying to take the lessons from today's readings and lecture as things to keep in mind when designing future projects, not as hard and fast guidelines for how information "should" be presented.

== Matt Gedigian - Dec 02, 2008 12:24:07 am ==

@Nivay Here is a page with Apple's [http://developer.apple.com/documentation/UserExperience/Conceptual/AppleHIGuidelines/XHIGIcons/chapter_15_section_7.html#//apple_ref/doc/uid/20000967-TPXREF106 guidelines for icons], which specify viewing and lighting angles.

Here is a video of some of Adobe's experimental [http://hothardware.com/News/Adobes-Interactive-VideoObject-Manipulation/ video manipulation] tools.  Some of them are similar to what Maneesh demo'd, but are being used for video rather than being combined into a single image.

== Ketrina Yim - Dec 02, 2008 12:12:14 am ==

Before this lecture, I never really thought about the considerations technical illustrators made with regards to lighting. It just didn't occur to me that the lighting in an illustration was a fabrication designed to reveal the most detail without all the "inconveniences" of realistic lighting (mainly cast shadows from one part of an object obscuring other parts). Shaded technical illustrations always seemed to "look right", so I didn't feel the need to question the realism of them. This probably comes from me already knowing that they are visualizations and that only the accuracy of the image in capturing detail matters.

== Michael So - Dec 03, 2008 11:36:19 am ==

It is pretty interesting and sort of scary the impact shading has on interpreting three-dimensional shapes.  From that demonstration with determining which object appeared convex versus compared, these interpretations can change by easily turning the images of upside down.  I wonder if there's a way to make these interpretations permanent no matter at what orientation a person sees the image.

On the Perceiving Shape From Shading reading, the conclusion about how the visual system uses simple rules to determine shape from shading, one being the single light rule.  But what if the scene has more than one light; does a person still perceive the scene as having one light source?  Or perhaps the person does not really think about it because I myself haven't really looked at something and pondered how many light sources there are.

== Calvin Ardi - Dec 04, 2008 01:26:23 am ==

I have the same (or similar) question as Michael: do people with the untrained eye (e.g., not someone who has been trained in photography and lighting) notice in photos multiple sources of light or that the image might be a composite of other images ([http://vis.berkeley.edu/courses/cs294-10-fa08/WWW/lectures-WWW/shape-shading/walk048.html example] presented in lecture). What seems particularly interesting is that similar lighting is done in presentations. The most obvious example coming to mind is at museums; if an object similar to the one presented in [http://vis.berkeley.edu/courses/cs294-10-fa08/WWW/lectures-WWW/shape-shading/walk045.html lecture] was being displayed, the lighting and angles would be carefully chosen so that people see something more like the right image as opposed to the one on the left. In particular, however, things seem a bit more off when looking at a picture that contains only one light source as opposed to something that may have multiple light sources; it's perhaps as if we expect to see these images and photographs to be perfect for viewing all the different angles and details.

== Witton Chou - Dec 07, 2008 02:48:12 pm ==

It is really interesting how the use of different shading and lighting techniques can help bring out or mask various features of an object. Often times, we look at pictures or models or drawings and our mind quickly interprets shapes and textures from them. What we often overlook are things like edge stroke and multiple light sources. While the pictures and representations may look normal, they are not always natural and often times our mind will ignore the details as long as the representation is clear. 

Hollywood uses a three point lighting technique consisting of a key light, a fill light, and a back light. Although we can clearly notice the key light, the principal light source, we often disregard the other two sources by assuming a single light source. However, if we were to take out the fill light and the back light, the picture would appear quite abnormal.

==[[User:Yuta Morimoto|Yuta Morimoto]]==
I engaged in the topic of the source of light.
I did not think that a brain could accept the only one source of light for entire image.
So, people probably presume that the direction of light is unique and never changed in the same image.
Conversely, the direction of light also may provide people with a clue to obtain the orientation of figures.
I guess that  consistent shading is a good way to obtain the orientation or direction of something.
Also, I  think that in the context of UI, shading provides information on the contour of a figure.
Shading and shadowing makes the edge of the figure more distinguishable and readable.
Especially, a simple figure such as recangle, circle or triangle becomes pretty much distinguishable when added a bit gradient or shadow.
However,  I have not seen a great example taking advantage of the effect of shading and shadowing in UI.
I found that they help us to recognize a contour, but do not help to obtain other features.

== HeatherDolan - Dec 08, 2008 12:03:49 pm ==

This lecture and the readings made me recall a paper/presentation I saw a few years ago (possibly written in by someone in Berkeley's vision science program, I wasn't able to find it) about how much information about an object and/or environment humans extract from shadows.  We don't pay much attention to shadow in the "real world", but if you look at drawing, painting, photography etc. there has been a huge focus on lighting and shadow for a long time. 

Ramachandran's work is really interesting.  The different responses in class to various convex/concave visualizations were fascinating.  My perceptions flipped back and forth frequently.  It was sort of strange but interesting to be cognizant of the fact that our brain is poised and ready to process this type of information all the time without us even being aware of it.

== David Poll - Dec 08, 2008 02:37:56 pm ==

Well, that discussion on how shading makes one interpret 3-dimensionality made me feel like a freak of nature :).  Nonetheless, I found it fascinating just how much of a difference it makes.  I'd be curious to know, however, if there was anyone else out there who got the "alternating" effect that I did (maybe we have something in common!).  It's quite possible, though I don't know exactly how it would be connected, that I experience the effect because I have a proprioception disorder, a disorder in which my muscles don't tell my brain how they are oriented in space.  I wonder if that has skewed some of my ability to perceive 3-d renderings.  I know I have always had difficulty making sense of spatial problems anyway, so maybe they're related?

== Seth Horrigan - Dec 08, 2008 02:29:53 pm ==

Having seen many hand-crafted illustrations, I was familiar with ways in which they differ from actual photographs, but this was the first discussion in which I was engaged covering ''how'' those differences are achieved. As Calvin mentioned, I have noticed cases where very special lighting is used to reveal key features in museums; however, the inverse is often more noticeable - when the museum curators do not pay adequate attention to the lighting of the display and key features are hard to distinguish or completely invisible.

What I found most interesting in Maneesh's relighting work was the ability to achieve nearly impossible lighting effects. Erasing shadows, highlighting edges, illuminating key angles. Certainly with just the right combination of light sources, that display could be provided in a single photograph, but the interface provides a much more friendly, robust manner of achieving that. Sadly, it does require quite a bit of finagling with the light sources to achieve the initial photographs, but far less than would be required to simultaneously light the scene as needed. I do have to wonder though; is it easier to achieve the "best" rendition using computer systems and lighting tools, or to be able to physically move around a bunch of lights. I would think the former, but often times the act of moving physical objects can lead to quicker and better results.

Lastly, like Heather, I really liked Ramachandran's work on perception. I will definitely share the hollow masks and the perception tricks with friends over the Christmas break. I love having those sort of things so that when somebody asks, "what classes did you like most last semester", I can demonstrate some of the really engaging properties of my courses.

== Maxwell Pretzlav - Dec 08, 2008 02:49:00 pm ==

I found the Ramachandran article really fascinating.  Despite our readings on pre-attentive visualization, I had never really thought about it in terms of 3D geometry and recognition of 3D shape.  I found the symmetry example particularly fascinating — my mind's view of the spheres making a symmetrical pattern was so strong that I didn't even understand what other symmetry the article was referring to until I rotated the page and had the "ah-ha!" moment of realizing what my mind had actually been doing.  I was surprised to find that on the images with both concave and convex shapes I can get my mind to either see the spheres properly as some convex and some concave, or all spheres as convex (despite the impossibility of a light source causing that), like David claimed to see in class.

== Chris - Dec 09, 2008 12:41:24 pm ==

Regarding the psychological observations made in the Ramachandran paper, the illusions discussed have a clear benefit in giving the "most likely" interpretation of an object, given an image of it.  What would be perhaps more interesting than the "why" of these observations is the "how" of them.  For example, if one were to dynamically observe, say, the figure on page 77 (the concave face lit from above) with additional context, one would quickly infer the correct non-typical geometry of the object.  A question which interests me is what level of consciousness (in some sense) is responsible for this.  It would seem (and many of the psychological experiments referenced in the course suggest) that there is a level of vision which exists after pre-attentive observation and before conscious analysis.  This mid-level analysis is probably the most approachable at the moment from a computer vision perspective (the lower level being solved and the higher level being inaccessible).

== Simon Tan - Dec 12, 2008 02:13:31 am ==

The reading on Light Collages brings up an interesting point that I wasn't sure was addressed in lecture: The reason inconsistent light schemes for revealing geometry works well and isn't disconcerting to viewers is because most untrained humans do not notice the inconsistency. Is this because most existing art ignores the geometry of consistent lighting (as suggested by the paper) and we've gotten 'used to it', or is it because human vision is intrinsically weak at detecting lighting inconsistencies?

This reminded me of a problem I encountered during class: When we were looking at the use of shading spheres to automatically color surfaces on 3D models (Sloan), I wondered why I was not able to detect what was "wrong" with the result. Eventually, I had to just be told that it was because there should have been shadows that weren't there; it might be just me, but I really had no complaints about the result until it was revealed outright.

@Matt: Those Apple icon guidelines read like they belong to an earlier lecture about Perspective. Accidental views, et al. I was surprised at the guidelines, though, because they seemed to imply that it was all right to line up icons using different perspectives all in a single line on your Dock.