The Digital Michelangelo Project
Recent improvements in laser rangefinder technology, together with algorithms developed in our research group for processing range data, allow us to accurately digitize the external shape and reflectance of many physical objects. As a demonstration of this technology, a team of 30 faculty, staff, and students from Stanford University and the University of Washington spent the 1998-99 academic year in Italy digitizing the sculptures and architecture of Michelangelo. During our year abroad, we also scanned all 1,163 extant fragments of the Forma Urbis Romae, a giant marble map of ancient Rome.
One of the original aims of this digitization project was to create an archive of unusually large and detailed 3D scanned models for use by computer scientists, artists, art historians, and archeologists. However, as we have reported in our papers and talks, building these 3D models has proven more difficult than we expected. Part of the difficulty lies in the large size of our datasets - more than a billion range samples for the David. Another difficulty, discussed in our 2000 SIGGRAPH paper, was the challenge of deploying into the field a mechanical scanner that was accurate to 1/4mm accuracy over a working volume of 5 meters - a "geometric dynamic range" of 20,000:1. As a result, our scans are slightly warped - about 1mm for each 200mm of travel along the surface, or about 0.5%. To date we have produced a good model of St. Matthew at 0.29mm and decent models of the David at 1mm and 2mm (but see note immediately below). Our models of the other statues we scanned are distinctly inferior to these. In some cases, the raw data itself is relatively poor, such as our scan of the highly polished statue of Night.
Although our archive is incomplete, we have decided to make our models and our entire corpus of raw range data available to the scholarly community, with the hope it will spur other researchers to address the problems we have faced. Undoubtedly, these efforts will lead to new solutions for these problems. If they also lead to better models of these statues, we would be happy to incorporate these models into our archive, with suitable credit given.
Note added July 21, 2009: Although it has taken 10 years, we finally have a full-resolution 3D model of Michelangelo's David. It is a 0.25mm VRIP model of the entire 5-meter statue, totaling nearly a billion polygons. The model was aligned by Benedict Brown and Szymon Rusinkiewicz using the non-rigid alignment method described in their 2007 SIGGRAPH paper. The new model is not perfect, but neither was our scan of the statue. In particular, the model still contains small holes where our scanner could not reach. Nevertheless, this is now our largest model, and as of 2009 it may be the largest geometric model in existence of a scanned object.
Click on the logo above to view a complete catalog of our models and raw data.
See below for instructions on how to obtain permission to download the data.
The archive whose catalog is linked above contains our best 3D models of ten statues by Michelangelo: the David, the four Unfinished Slaves (Youthful, Bearded, Atlas, and Awakening), and St. Matthew, all located in the Galleria dell'Accademia in Florence, and the four allegorical statues (Night, Day, Dawn, and Dusk) from the tombs of Lorenzo di Medici and Giuliano di Medici, in the Medici Chapels. Click here for a poster showing images of everything we scanned. As of July 2009, our largest model (the David) contains 940 million polygons, and our second largest (the St. Matthew) contains 386,488,573 polygons. For those with a more limited appetite for polygons, the archive also contains simplified models of the David, created using Paolo Cignoni's OutOfCore simplif package. Finally, for those with no appetite for polygons, the archive contains input files for QSplat, our multiresolution point-based viewer for large polygon meshes.
In addition to 3D models, the archive contains the raw range data (about 10 billion range samples, comprising 2000 range images) from which these models were built, and files giving our best estimate of the relative alignment of these range images. Most of this range data was acquired using the Stanford Large Statue Scanner, fabricated for us by Cyberware of Monterey, California. For those hard-to-reach places (we didn't move the statues), we used a second scanner - a jointed digitizing arm and small triangulation laser rangefinder made by Faro Technologies and 3D Scanners Ltd. On the positive side, these datasets contain connectivity and line-of-sight information, as returned by our scanners. This makes them useful for testing alignment, surface reconstruction, and hole filling algorithms. On the negative side, this is real range data, so it contains miscalibration, noise, holes, grazing-angle scans, and other unavoidable artifacts of laser scanning.
Since the summer of 2004, we have done no further work on this archive - mainly for lack of funding and the time to seek additional funding. (However, see the note above on the fine work done by Benedict Brown and Szymon Rusinkiewicz in 2009 on the statue of David.) We welcome requests for licenses from any research group or institution that wishes to help us finish this monumental labor. Since scanning large objects at high resolution will always yield datasets with the artifacts enumerated above, developing principled methods for overcoming them seems like worthwhile research, and good solutions would undoubtedly be publishable.
In addition to range data, we also acquired 40,000 high-resolution color photographs of the statues, taken under calibrated illumination. Since many of the statues have been cleaned since we acquired these photographs, we consider them obsolete, and we have not included them in the archive. They also occupy hundreds of gigabytes, making it difficult to download them or even distribute them on DVD. However, we are willing to make this data available to scholars on request. Also not included in the archive are range scans and color photographs of two architectural interiors: the Tribune del David in the Galleria dell'Accademia - home of the David, and the New Sacristy in the Medici Chapels - home of the four allegories. The sacristy was designed by Michelangelo. This architectural data was captured using a prototype time-of-flight laser rangefinder made by Cyra Technologies. Finally, our archive does not include the large light field we captured of Night, nor our range and color data of the map fragments of the Forma Urbis Romae, which has its own licensing procedure.
Unless otherwise noted, the models in this archive are produced by our software pipeline for aligning and merging range images. Our merging algorithm is called VRIP. Some reconstructions are performed with space carving, some without; see our Siggraph 2000 paper for more details. The VRIP algorithm has been incorporated into a library called VripPack, which is available for download. This library is in turn invoked by the Scanalyze software package, which was used to create all of the models in this archive. Another software package that might be of interest is Volfill, our diffusion-based hole filler for large polygon meshes.
Aside from our small quantity of Modelmaker data (whose format is available upon request), our raw range data are represented using SD files (extension .sd), and our merged 3D models are represented using PLY files (extension .ply). Both are home-grown file formats. SD files encode range images as rectangular arrays of points. These files also contain metadata that describes the geometry of the range scanner used to acquire the data. This geometry is used by the Scanalyze package to derive line-of-sight information for various algorithms. PLY files describe a 3D model as a polygon mesh. Aside from some header information, the file contains a list of X, Y, and Z-coordinates for each triangle vertex in the mesh and a list of 3 vertex indices for each triangle. Source code for reading these files can be downloaded from http://graphics.stanford.edu/pub/zippack/ply-1.1.tar.Z". We also have a utility for converting PLY files to Inventor files. Click here to download the executable. Another site with information about PLY files is the PLY File Format page of the Georgia Institute of Technology's Large Geometric Models Archive. For more information about the structure and format of the files in our archive, click on this help file.
Our contract with the Italian authorities, which includes the statement of rights and limitations reproduced at the bottom of this web page, permits us to distribute these models and range datasets to established scholars, for non-commercial use only, and it prohibits us from placing them "on the Internet", i.e. for automatic download. To comply with these restrictions, we ask any scholar who would like to use this data to send us email at dmich-request at graphics dot stanford dot edu stating:
The rights to copy, distribute, and use the 3D computer models and range data (henceforth called "data") you are being given access to are under the joint control of Marc Levoy, director of the Digital Michelangelo Project of Stanford University, and the Soprintendenza ai beni artistici e storici per le province di Firenze, Pistoia, e Prato. You are hereby given permission to copy this data in electronic or hardcopy form for your own scientific use and to distribute it for scientific use to colleagues within your research group. Inclusion of rendered images or video made from this data in a scholarly publication (printed or electronic) is also permitted. In this case, credit must be given to the Digital Michelangelo Project, Stanford University. However, the data may not be included in the electronic version of a scholarly publication, nor placed on the Internet. These restrictions apply to any representations (other than images or video) derived from the data, including but not limited to simplifications, remeshings, and the fitting of smooth surfaces. The making of physical replicas of the artifacts represented by this data is also prohibited, and the data may not be distributed to students in connection with a class. For any other use, including distribution outside your research group, written permission is required from Marc Levoy. Any commercial use also requires written permission from the Soprintendenza. Commercial use includes but is not limited to sale of the data, derivatives, replicas, images, or video, inclusion in a product for sale, or inclusion in advertisements (printed or electronic), on commercially-oriented web sites, or in trade shows.
If your status seem legitimate, your intended use appropriate, and you accept the terms of licensing, then we will set up an account on our server that permits you to download the models and range data from our archive. Your email address will also be added to a licensee mailing list. Among other purposes, we will use this list to announce the addition of new models to the archive.
Since announcing the availability of our 3D models and range data, we have been inundated with requests for licenses. In particular, we receive many requests whose legitimacy we cannot easily judge, and we are concerned about our data being used inappropriately. Our short-term solution is to limit distribution to: (1) published researchers who are currently affiliated with a university, industrial research laboratory, or other major institution, or (2) scholars in museums or other cultural institutions. To permit students to obtain access to the models, we ask them to join forces with their academic advisors. In other words, students may not apply for licenses; their advisor should submit an application on their behalf. For scholars working in academic or industrial research laboratories, their lab director should apply on their behalf. This rule applies also to postdoctoral scholars. For the time being, we are strictly enforcing these restrictions. If you do not satisfy them, please don't bother applying for a license; we won't respond. Note also that the data cannot be distributed to students in connection with a class, and the making of physical replicas from the data is not allowed.
Our long-term solution is to turn licensing of this data over to a third party, for example a non-profit intellectual property management organization, to provide comprehensive long-term archiving, licensing, and distribution of our archive. We have also developed ScanView, a secure viewer that permits unlicensed users to examine our 3D models, but not extract the underlying data. If you want to look at our models, but you don't think you qualify for a license, this is the link for you.
As you browse our catalog and think about how you might use our 3D models and range datasets, please remember that the artifacts they represent are the proud artistic patrimony of Italy. Keep your renderings and other uses of the data in good taste. Don't animate or morph the statues, don't apply Boolean operators to them, add silly hats to them, or simulate nasty things happening to them (like breaking, exploding, melting, etc.). Choose another model for these sorts of experiments. Also, exercise reasonable caution to prevent the data from wandering beyond your research group. Jointly, our good behavior in this matter will encourage other cultural institutions to permit their 3D artifacts to be digitized and made freely available for scholarly use. Finally, if you plan to use our models to test a surface reconstruction algorithm, see our caveat about range data versus 3D models.
If you are interested in a commercial use of our data, you must obtain permission from the Superintendency of Fine Arts in Florence and from Stanford University. This process typically begins by sending a letter or fax to the director of the Galleria dell'Accademia, and an email to Marc Levoy, director of the Digital Michelangelo Project. Depending on your intended use, you may be required to pay royalties. Note: In the special case of reproducing small numbers of images in a technical book, it usually suffices to send an email requesting permission for such use to Marc Levoy.
Note added April 19, 2021: We have given a copy of our archive to Stanford Libraries for safe keeping long term. We have also been working with Italian authorities to negotiate a new contract that will permit Stanford Libraries to issue licenses on their own authority. However, these negotiations have not yet borne fruit. Therefore, for the time being Marc Levoy is still the only authorized source of licenses.
Notizia: Questi modelli elaborati al computer, immagini computerizzate, e fotografiche sono proprietÓ del Progetto Digitale Michelangelo e la Soprintendenza Per I Beni Artistici e Storici per le Province di Firenze, Pistoia e Prato. Non possono essere copiati, scaricati da internet su un file, inviati, o riprodotti in nessuna forma, incluso la posta elettronica o il web, da nessuna persona per nessun motivo, senza un permesso scritto da Marc Levoy, il direttore del progetto. Eventuali usi commerciali esigono anche il permesso scritto dalla Soprintendenza.
Clarifications: The prohibition in the preceding paragraph against copying does not apply, of course, to the ephemeral downloading and copying associated with browsing these web pages. Also, permission is hereby given to reuse the project's logo image in its entirety, even to incorporate it into web pages, but only as a clearly labeled representation of the Digital Michelangelo Project. All other use requires the express permission of the project director.