Imagine a search engine capable of recognising the details of paintings rather than words. This is the challenge facing a joint Microsoft/INRIA research project to recreate, in 3D virtual form, a two thousand year old house in Pompeii.
Computerised excavations in Pompeii
Two hundred years ago, a group of travellers discovered the highly adorned walls of Queen Caroline's House, a Roman aristocrat's residence in Pompeii that had been buried and, over the centuries forgotten, under the ashes of the great volcano eruption in AD79. The wealth of decorative art included delicate paintings, rich with colour. In the intervening decades since the house was rediscovered, however, the splendid finds have been damaged by rain, wind and light, leaving bare and eroded grey walls.
Today, researchers are attempting to create a 3D virtual model of the Roman-era mansion, named later, in the nineteenth century, for Caroline Bonaparte, a former Queen of Naples.
Pompeii, Queen Caroline’s House: view of the atrium in 2006.
This is just one of the goals of the ‘Images and Videos Mining for Sciences and Humanities' project led by the joint research centre set up by Microsoft and the French National Research Institute for Computer Science and Applied Mathematic (INRIA). This centre, launched in early 2007, employs around 30 researchers in Saclay, south of Paris.
The Centre's objective is to pursue fundamental, long-term research in formal methods, software security and the application of computer science research to the other sciences. In addition to archaeology, the ‘Images and Video Mining' project includes applications in the realms of the environment and the sociological study of the behaviour of characters in films or television advertisements.
The 3D reconstitution of a Roman mansion in Pompeii brings together computer scientists and archaeologists. To begin with, the project team has to collect about two hundred engravings, sketches, paintings and photographs made in the two hundred years since the building was unearthed from the ashes. Once correlated, these historical documents will help to bring the opulent residence back to life, just as it was at the time of its re-discovery.
Pompeii, Queen Caroline’s House: view of the atrium circa 1810 after F. Mazois, Pompeii’s ruins: second part.
This collection of icons will also see the development of a new computerised visual recognition application. This emerging technology is designed to automatically recognise a fragment of an image representing an architectural detail or a ceramic in a database.
“We are making our own computerised excavations,” explains project leader Jean Ponce, a former researcher at the University of Illinois at Urbana- Champaign, who specialises in Computer Vision, Computer Graphics and Robotics. Jean Ponce currently heads the Willow research team, a joint project between INRIA, the Ecole Normale Supérieure (ENS) and the Centre National de la Recherche Scientifique (CNRS).
“We have three years to test a study protocol that is already applied to photographs, but not to paintings,” adds Jean Ponce. “One of the greatest challenges lies in recognising objects whose appearance differs slightly from one document to another, especially in terms of the proportions. And sometimes, the artists add more details.”
The goal is to locate the detail of an image in a multitude of archives in just a few clicks: an achievement that would revolutionise archaeological practices.
The researchers are planning to build a search engine in order to test the Pompeii, Queen Caroline's House: view of the atrium circa 1810 after F. Mazois, Pompeii's ruins: second part. visual recognition technology. The goal is to locate the detail of an image in a multitude of archives in just a few clicks: an achievement that would revolutionise archaeological practices. “Currently, it can take months of research to find a document. And we have to remember that such and such a sketch corresponds to the same wall painting as the one we saw a few weeks earlier,” explains Hélène Dessales, lecturer in archaeology at the ENS and a member of the project team.
By comparing the images, the researchers will also gain a better understanding of the working methods of ancient architects and their 19th century counterparts, and of how buildings decay over time depending on the climate and diseases affecting stone. “A computerised tool will greatly help us to take decisions on policies to conserve and preserve our heritage,” says Hélène Dessales.