The Six Faces of 3D
In a previous design tip for creating models using 3DTin, I briefly mentioned a low-tech hack for gridding up a source image in Inkscape so that you can more easily eye-ball what you are doing within your 3D design app. I feel it is worth mentioning this again to show how I’m using this technique right now for something far less 2.5D than the Mario Cloud pendant I was making at the time.
“Gridding up” an image (ie marking up an image with a ruler or line tool) for orthographic, 1-pt, 2-pt, or 3-pt perspective offers a designer a tool for analyzing cues for depth, profile, and placement: a rather old trick frequently used when generating technical and architectural drawings where the measurement of component parts is the key information. Gridding plays a particularly important role in computer visualization where moving the camera position within a modeling environment is now possible (even probable). How will the model skew when seen from this or that angle?
While there has been a lot of discussion lately about the need for an accurate, inexpensive 3D object scanner priced proportionate to a MakerBot printer, here is a low-tech technique that will help you get to work modeling while other folks are pouring their energies into clever solves for single-angle point-cloud generation or the construction of elaborate tracking/capture rigs.
When you model an object from real world you can move the camera (yourself) around the object and see it from multiple perspectives. (Or turn it around in your hand.) While many 3D modeling apps allow you to move the camera anywhere around a model both possible and impossible in the real world, these apps also tend to have options built-in to lock your camera perspective to right angles to the model on the X, Y, and Z-axes. This means that you can typically use quick key commands to move to each of the “six views” (think of a 6-sided die) in turn: X+, X-, Y+, Y-, Z+, Z-.
Well, rather than pulling out a protractor, a graphic calculator, and working out the math for gridding up a photo from a single perspective so that it contains all of the information you need to model it, you have the luxury of capturing the six default angles in the real world extremely simply (ie straight-on with much less attention to skew) and marking over the image an orthographic grid and the additional notes you need when considering your model from that locked view. By using your reference sheets and switching between the six angles, you can very quickly create a baseline for even an incredible complex model. And once you have a model that fits essentially to each view, you can better put your efforts to the tweaks, deformations, texturing, carving, and other techniques that will help you reach your goal for the project.
Centimeter Gridded Coffee Mug (Annotated)
And how is this for a low budget 3D scanner? If your object is small and not terribly heavy, you can place it right onto the bed of a scanner or photocopier and grab a photo of each of the six views. To instantly generate a printable grid in Gimp go to Filter > Render > Pattern > Grid for a really powerful/tunable grid filter. Is your shape not a cube? No problem! Use bags of dried beans, string, or other no-fi support to position the model so that it is where you want it. Did you lose part of the object in a drawing due to support? No problem! You can write down or trace over the image with any additional information you need that is obscured from that angle.
If you have a more permanent place to setup a rig, you might try setting up a camera at a fixed height over a table and grab your references this way. As the topmost surface of the object will likely be at a different height from side to side, you might take a queue from Marty and Amy’s Nickel for Scale project, placing a common object such as a nickel on the top of the object you are referencing so that you can rescale each perspective reference to match the known dimensions of the reference object to adjust for the differences in height.
This method is not perfect — i.e. it still depends on your skills as a modeler to move from the simple shape to the completed part — but it has a nice high-proportion of useful data to noise that can actually be an advantage over analytic scanning tools that have difficulty differentiating what details are important for the model. What is the essential form and what is texture or light scatter? Your eye will tell you far more quickly than your tracking/scanning algorithm. When you get comfortable with the six-views method you can go from zero to basic model fast enough that you can target your modeling attention to specifically what you want out of the model instead of fighting with simplifying a messy points cloud.