photographist

“I was interested in programming tasks to be performed by a computer and quickly developed many little applications during my studies. The aim of computer vision is to achieve with a computer the results of the complex operations carried out by the eye and the brain in human vision. It involves analysing images with a view to extracting information that can be used to identify the content of these images. That was the subject of my doctoral research in image analysis. I continued with these subjects for my post-doctoral research at INRIA, applying them to large volumes of images. We speak of image indexing. The idea is to be able to find information automatically from large volumes of images, such as those on Internet, without using keywords. I was then hired as lecturer and researcher at CNAM to implement this type of analysis and image indexing.

Now as director of research at IGN, I apply these notions of analysis and image indexing to street view data and to overhead and satellite imaging. The institute's missions are to produce data to scan the entire French territory. Entering this information creates vast volumes of data to be managed. For the city of Paris, for example, scanning the images relating to every street produces billions and billions of visual characteristics. These need to be managed and the information has to be able to be found. Given the volumes at play, we try to get the computers to extract the visual information – detect objects, places, faces, etc. – so as to be able to organise, manage and also navigate through all of this data in a more automatic way. This can lead to applications which automatically detect forest fires or traffic congestion.”

next