(This post is loosely based on a short presentation I gave in the lunchtime public engagement workshop “Classics and History in 3D” at the Being Human Festival here in Senate House on November 22 last year.)
For this exercise, I wanted to walk through the process of taking a Greek vase (in this case a miniature kylix from the Ehrenberg bequest in the ICS/Hellenic and Roman Library), scanning it as a digital 3D file, cleaning up the scanned image, and printing it again using the Institute’s CraftBot 2 3D printer, bought last year as part of a package of 3D kit for experimenting, learning and teaching purposes.
I started by choosing the small kylix, both because it is a relatively simple shape, the inside is visible (making a clean scan and a print possible), and its small size should make a fairly quick print (although more on that later!). I imaged the kylix using a scanning technique called photogrammetry, which is both easy and cheap: the only hardware needed is an entry-level camera—my three year-old cellphone was more than adequate—and there is free software available for processing the images into a 3D model (although we used Agisoft Photoscan, which costs $59 for an academic license).
Photogrammetry works basically by taking a series of photographs of an object from every possible angle. It is important for light and position to be completely consistent between the photographs, and it helps to have an irregular background (such as an illustrated page of newsprint) to help the software tell between similar-looking areas of a repetitive or symmetrical object. So long as each point on the surface of the object appears in at least two clear photographs, it should be possible to create a good model. I took about 120 shots to be on the safe side, and processed about 100 of them after removing out-of-focus or overexposed images.
Running photogrammetry software on the photographs takes anything from 5 minutes or so for a medium-quality 3D model, to up to several hours for the highest quality, using hundreds of high resolution images. Mine took about 10 minutes total, going through all the stages:
- Aligning images (which leaves you with a visualisation of the position from which each photograph was taken, like this one);
- Building the dense cloud (calculating from these images a series of points in 3D space on the surface of the object);
- Building the mesh (turning the surface of the calculated 3D geometry into a series of triangular faces);
- I skipped the fourth step, which would normally be to build the texture for the 3D object (i.e. use fragments of photographs to give each face of the model the appropriate colour and shade), because I only want this model to 3D print, and that doesn’t use colour or texture.
The next step was to edit the 3D model, in a piece of software like Meshmixer (currently free), to remove any extraneous pieces, such as part of the tabletop or other background objects, to smooth some of the edges, patch any holes in the image, and turn it into a “watertight” 3D solid, ready for printing. At this point I found it useful to artificially flatten both the base and top edge of the vase, by slicing off the irregular surface that was caused by the low quality of the 3D scan. This again makes it easier to extrude the model onto the flat bed of the 3D printer.
The final software I used to prepare the 3D object was CraftWare, the custom tool that comes with the CraftBot printer range. This is used to scale the object (I set its longest dimension to 107mm, the same as the original vase), to set the position and size of supports—which are needed for those parts of the object that overhang, and which could not be printed using additive layers with nothing beneath them. For a complex shape, such as this one with three areas of support columns, it also helps to print a “raft” beneath the whole object, to make it more stable and less likely to detach from the base mid-print. This is also the point to choose a print quality, ranging from lowest (which will take a little under two hours to print an object of this size and complexity), to the highest, which is slower and uses finer layers of filament, and can take upwards of four hours, even for this small vase. The object shown here was printed at the highest quality.
When printing the object, there are a few steps to keep in mind, including keeping the base plate clean and aligned to the extruder nozzle, which can help make a print more likely to succeed. When we first got the printer, our inexperience led to about one in three attempt prints to detach or fail in some other way. This seldom happens now that we know the precautions to take.
We ended the Being Human event with a short discussion of the obvious inadequacies of a 3D printed object (the weight, texture and colour and inaccurate; surface design is absent; it doesn’t have good tactile or haptic resemblance to an ancient vase; there are usually rough areas on the object—such as the underside and beneath the handles, in this case, where supports were printed and filed off).
The audience then offered some suggestions on advantages of the 3D object over the original artefact: we can hand them around a classroom without concern they might be damaged (I tossed one across the room for an audience member to catch, to prove this point); we can print multiple copies (I had three with me, not counting the two other, inferior copies I keep to show lessons we learned about printing); we can experience them in ways we cannot the original—I could pour water (or wine!) into a printed kylix and see how it feels as a drinking vessel; we could experiment with painting or otherwise decorating the object in the ancient style; we can reconstruct missing parts of an object, whether we know exactly how they would look, or have to guess.
I was pleasantly surprised by the richness of suggestions the audience came up with, and the enthusiasm they showed for even this modest example of 3D printed replicas of ancient artefacts.
The 3D printers and other scanning/modelling gear that the ICS and IHR purchased last year are here to encourage learning, experimenting and collaboration. If you have any ideas for using 3D technologies in researching, teaching, communication or engagement, please do get in touch, we’d love to talk it over with you.
by Gabriel Bodard