Level of Detail

With the 2024 SketchUp 3D Basecamp conference fast approaching, I’ve been thinking back to some of the most unexpected and exciting things we learned at the last conference, back in September 2022. Not only did we discover new tools and techniques, but also some applications for 3D modeling as niche as our own. I was particularly curious about Matthew Nicholls’s presentation “SketchUp for Cultural Heritage and Education” based on my own interests, and discovered that the questions Nicholls was dealing with in his work were very closely tied to our own at Fat Pencil.

Nicholls, a professor of classics at the University of Reading and senior tutor at St. John's College, University of Oxford, has been creating an extensive 3D model of ancient Rome for over a decade. The model's aim is to give students and other viewers an accessible way to explore the city circa 315 AD, with the option to view it from any angle, as a whole or by individual streets and buildings, and ultimately gain a better understanding of urban life in the Roman Empire.

This still was taken from the video above, portraying a small portion of the 3D model -- though the architectural details are carefully rendered, not much focus is given to defining the sculptures.

To balance the huge amount of information with available time, computer memory, and the larger picture, Nicholls had to create some parameters for his work. The image above illustrates the kind of choices Nicholls was faced with: time and computer memory is saved, for example, by not making the horse sculptures in the image above particularly detailed.

By definition, all models and maps are partial representations, because their creators are always forced to make decisions about details - what to include, what to highlight, and what to leave out altogether. For physical objects, having a particular scale helps determine level of detail, but for digital models, level of detail is based on somewhat different considerations and limitations.

Level of Detail (LOD) is a specific concept in computer graphics, referring to the complexity of a 3D model representation. It is most often discussed in the context of video game modeling or architecture, but is useful in other fields as well. For example, two SketchUp Basecamp presenters from urban planning firm Houseal Lavigne shared a slide describing their LOD rating system. (Scroll to fig. 4 in this article).

If you're keeping track, that's two mentions of LOD in one conference, and it got me thinking about how the concept applies to Fat Pencil’s work. Here's an example, using the historic Couch School (now called Metropolitan Learning Center) in Portland, modeled at three different Levels of Detail.

The images above represent what Low, Medium, and High Levels of Detail typically look like in our studio. Although we don't have hard-and-fast rules for what each LOD includes, it's a helpful frame of reference for getting on the same page, both within our team and with our clients.

Our early conversations about a new project always include a discussion of objectives and audience. In order to create the most effective 3D model we can, we focus on the crux of the story and what elements are needed to communicate it. This can mean that we only model furniture in the rooms where an alleged crime occurred, or only put panel details on a door if a bullet was shot through it. For projects where our goal is to illustrate multiple phases for a construction logistics project, we may add realistic textures and figures to give viewers a better sense of how the finished project will look. Our eyes tend to be drawn toward more detail and color, so LOD also helps direct the audience's focus.

For the case we call the NE Portland Shootout, we created a model that contains a range of LODs. Since the incident occurred in the street, all the surrounding buildings are modeled at a low LOD, while most of the vehicles, which were closer to the action and constrain the sightlines, are at a medium LOD. The vehicle of the victim, however, is significantly more detailed and precise, since his actions would have been restricted by things like the height of the car and the size of the door opening.

This shootout example also illustrates another important factor in determining LOD: the source information available. In this case, we had access to a laser scan of the site of the incident, so we could take measurements of the vehicle and make sure our model was accurate to the inch. Other parts of the model, like the buildings, would not change the story if they were only accurate within 6"-12", so we did not spend time measuring them in the laser scan, relying instead on our aerial mapping program data.

Measuring the laser scan allowed us to be certain of the vital dimensions of the vehicle in question.

The question of source information is a significant one: it's possible that part of the reason Professor Nicholls did not model the details of the sculptures in the image I called out is that the information he was relying on for that building was limited, and the decorative details are more suggestions based on clues than specific objects that undoubtedly existed as modeled. Another idea that I appreciated in the Basecamp presentation was about "conventions for showing doubt," which we'll take a look at in the context of Fat Pencil's work in a future post.

We always want to use our time, and that of our clients, wisely, so we don't waste attention on model elements that will never been seen or aren't vital to the story or aim. Our graphics are meant to look good, but they are first and foremost tools meant to facilitate communication and understanding, so when we open up SketchUp with a sense of the LOD required, we can focus on the big picture -- it's all part of the "Fat Pencil approach."

Alexandra Friedman is a Designer at Fat Pencil Studio