Issue link: https://digital.copcomm.com/i/786150
28 cgw j a n u a r y . f e b r u a r y 2 0 1 7 The point cloud (the 18 billion points) has to be made into a mesh (that is, the points are joined to form triangles). The triangles furnish the surface onto which textures can be generated so that the sur- face appearance of the 3D model is quite close to the original statue. Typically, the textured 3D model has much more data than is needed for visual- ization on a Web page, so the final step in the process entails decimation of the model to about one million to two million triangles. "We keep the full, undecimated model to be used when a scholar or museum conser- vator needs very detailed information about the statue, like when the person wants to be able to see tool marks, scratches, or nicks on the surface," says Frischer. At 500 36-megapixel photographs per object, and with about 1,250 objects con- sisting of Greek and Roman statues, inscrip- tions, sarcophagi, altars, and reliefs, there is a massive amount of data to process. T H E C H A L L E N G E O F S C U L P T U R E Many of the statues are displayed against walls. This causes a problem for data cap- ture since the 3D models must contain all sides, including the back. In the project's first year (May through June 2016), Frischer's team – comprising four doctoral students as well as Cristiana Barandoni, the project coordinator based in Florence – concentrat- ed on sculpture exhibited in open areas. In the second through fih years, the remaining statues will be moved away from the walls by at least two meters so they can be easily photographed from all angles. This will be time-consuming and costly – one cu- bic meter of marble weighs 2.7 tons. The team's partner at the Po- litecnico di Milano is currently experi- menting with low- ering a high-quality, albeit very small camera with a wide-angle lens, behind the statues that are situated only a few inches from the walls. The initial results are promising, but experimentation with this approach is still in progress. Another difficulty the team faces is access; they can only work in the Uffizi on Mondays, when it is closed to the public, and they can only afford to be present in Italy six weeks in the summer. This slows down the pace of the work and explains why the entire project will take five years to complete. To mitigate this problem, they have involved Professor Gabriele Guidi and his researchers in reverse engineering at the Po- litecnico di Milano, which is a short 90-min- ute bullet train ride away from Florence. "Professor Guidi's group can work in the Uffizi many more Mondays than can our team at Indiana University," says Frischer. "We also trained some post-doctoral students of Professor Paolo Liverani at the University of Florence who have taken charge of digitizing the fragments in the storerooms. They can access them five days a week." Technicians edit the images, and the final result is then optimized and uploaded to a Web service, which then generates an embedded code that is added to the project webpage. H A U L I N G H A R D W A R E The laboratory is based in Bloomington, Indiana, but the project itself is in Florence, a continent away. To this end, another challenge the group faced was getting the computing speed required to process these huge images on-site. "We needed the computer power of a tower but didn't want the hassle, expense, or inconvenience of finding a special suit- case for it or dealing with the customs pro- cess," says Frischer. "And there was no way we could get enough processing or video memory in something so small as a laptop. The only solution available was the powerful but still portable Shuttle SZ170R8." The Shuttle's small form factor provides them with all the key components they need to do their work, especially the possibility of using the latest video cards, without all the bulk and weight that a tower entails. The Shuttle gave the group mobility without having to sacrifice computing power. The team also works with Indiana Universi- ty's Information Technology Services (UITS) to port the photogrammetric soware to a high-performance computing platform, which allows them to speed up processing time and handle more objects at the same time. The data is collected in Italy and FTP'd to Indiana University for backup and storage. The entire collection is expected to be processed, uploaded, and available on Indi- ana University's website by 2020. W H A T ' S N E X T ? While there is still a great deal of work to be done, Frischer is optimistic: Through this project at one of the world's most visited and respected museums, the word will quickly get out that they have now entered what he likes to call "The Golden Age of 3D" when it comes to digitization of three-di- mensional objects such as statues, archaeo- logical sites, and buildings. Compared to just five or six years ago, costs have decreased dramatically, process- ing time has been greatly reduced, and the results can be easily embedded on a Web page or used as an asset on the new VR and AR devices. "The future is bright. The interactive 3D model of a work of art or architecture will serve the same purpose in the 21st cen- tury as did the 35mm slide in the 20th," Fischer says. ¡ Nancy Napurski is a technology writer and has worked in CG for many years.