Myriam Servières

My work is based on the scale of the pedestrian moving through open urban space and focuses on the connections that this point of view enables between the real world and a digital double of the city. This digital double provides information that can be displayed in the real world. Conversely, capturing the real world from a pedestrian's point of view enriches or even allows to simulate a digital double of the city. The main goal of my research was to develop a digital mediation system that uses urban data for a pedestrian immersed in the city. To display information on site, Augmented Reality (AR) seems to me to be one of the most suitable tools to enable this mediation. AR is based on reality and adds digital information that the user can interact with. To realize an AR mediation tool with geolocalized data on site, the crucial element is the location of the pedestrian and, more precisely, the calculation of the pose (position and orientation) of the camera he is carrying. This geolocalization can be seen as a feedback loop with spatial data. The spatialized data of the city, especially the geometric ones, are an essential input for obtaining a geolocation, and once this geolocation is obtained, other data coming from the information extraction of images taken in the field will enrich these spatial databases. This part of my research lies at the intersection of geomatics and computer vision. This work lies at the crossroads of CNU sections 27 and 61, where I am both processing imagery to obtain location information cross-referenced to external sources and working with inertial signals. Most of my contributions deal with the problem of fine on-site pedestrian localization, with different input sources (from pure images alone to images augmented with 2D, then 2.5D and 3D GIS data as well as inertial data). In my further work, I extend this study of pedestrians' interaction with open urban space by using other modalities on the continuum between realityand virtuality. Evaluating the perception of urban environments through digital technologies requires the development of an experimental approach ranging from a fully controlled environment in virtual reality to an uncontrolled environment on the ground. Recently, I have been interested in the perceptual representation of environmental effects in virtual urban worlds. In particular, I have been working on the perception of wind in virtual reality in an urban environment. Finally, to qualify the perception of a pedestrian in an urban space, I am now working on deep learning qualification of the perception of a pedestrian in an urban space based on the analysis of 360° images of urban scenes. This interest in the automatic qualification of users' perception of urban spaces has led to a new research direction in my lab that deals with the visual perception and understanding of urban scenes in the context of qualifying the cities walkability.