Section: Software

Software packages

TransforMesh: Mesh evolution with applications to dense surface reconstruction

We completed the development of TransforMesh, started in 2007. It is a mesh-evolution software developed within the thesis of Andrei Zaharescu and recently submitted for journal publication [57] . It is a provably correct mesh-based surface evolution method. It is able to handle topological changes and self-intersections without imposing any mesh sampling constraints. The exact mesh geometry is preserved throughout, except for the self-intersection areas. Typical applications, including mesh morphing and 3-D reconstruction using variational methods, are currently handled. TransforMesh will is available as open source with LGPL on http://mvviewer.gforge.inria.fr/

3D feature detector (MeshDOG) and descriptor (MeshHOG) for uniformly triangulated meshes

This is a C++ implementation of a 3D feature detector (MeshDOG) and a 3D feature descriptor (MeshHOG) for uniformly triangulated meshes, invariant to changes in rotation, translation, and scale. The descriptor is able to capture the local geometric and/or photometric properties in a succinct fashion. Moreover, the method is defined generically for any scalar function, e.g., local curvature. Typical applications include mesh description for “bag-of-features” kind of representations, mesh matching and mesh tracking. Both MeshDOG and MeshHOG are available as open source with LGPL on http://mvviewer.gforge.inria.fr/ .

Shape and discrete-surface registration based on spectral graph matching

We started to develop a software package that registers shapes based on either their volumetric (voxels) or surface (meshes) representations. The software implements a spectral graph matching method combined with non-linear dimensionality reduction and with rigid point registration, as described in several recent publications and in [53] as well as in the PhD thesis of Diana Mateus [11] . The SpecMatch software package is publicly available as open source with GPL on http://open-specmatch.gforge.inria.fr/index.php .

Real-time shape acquisition and visualization

This software can be paraphrased as from pixels to meshes . It is a complete package that takes as input uncompressed image sequences grabbed with synchronized cameras. The software typically handles between 8 and 20 HDTV cameras, i.e., 2 million pixels per image. The software calibrates the cameras, segments the images into foreground (silhouettes) and background, and converts the silhouettes into a 3D meshed surface. The latter is smoothed and visualized using an image-based rendering technique. Currently this software package is commercialized by our start-up company, 4D View Solutions (http://www.4dviews.com ). We continue to collaborate with this company. The latest version of the software is available for INRIA researchers and it runs on the GrImage platform.

Audio-visual localization of speakers

We developed a software package that uses binocular vision and binaural audition to spatially localize speakers. The software runs on the POPEYE platform (see above) and it has been developed in collaboration with the MISTIS project-team and with the Speech and Hearing group of the University of Sheffield. The software combines

stereo, interaural time difference, and expectation-maximization algorithms. It is developed within the European project POP.