Final Report Summary - NORDIA (Non-Rigid Shape Reconstruction and Deformation Analysis)
Ron Kimmel’s group has significantly advanced the field of shape processing and analysis in various fronts. The group at the Technion has been working in the fields of applied and fundamental theoretic research exploring old and new geometries and importing tools from theory into the applied shape analysis arena. Such geometries include equi-affine differential geometry, similarity differential geometry, full affine differential geometry, diffusion geometry and metric geometry. The main question the group is trying to answer is how shapes should be treated when attempting to compare and find the correspondence between them. In order to emphasise the importance of the theory and tools developed by Kimmel’s group we refer to a news item `Behind the Mask’ that has appeared in the media. In that project, Kimmel and his student Dr. Raviv were approached by pediatrician Prof. Amirav and were asked to help in the design of new inhalation masks for toddlers. The three dimensional structure of each baby’s face is unique and differs from an adult’s face. Until now, inhalation masks for babies were just a scaled down version of the adults’ masks. Raviv and Kimmel discovered that within the range of possible variations of the 3D facial structure, it is still possible to detect three main types of facial geometries. This process is called clustering, the categorization of data into clusters of shapes which feature relatively little changes. Each group was allocated a “median child,” the child with the face best representing his group, who functioned as a model for the mask design for each group. After choosing the representative child, they examined each group for the child who was most different (geometrically speaking) from the representative. They then inspected the computational adaptation of the mask to that child and made sure it was still suitable despite the variations. After careful examination and result analysis they used a 3D printer to ‘print’ the three median face types, and these were the models used for the mask design. In order to ensure that exceptional faces in each group also received proper care, they printed the face structure of the most different child in each group, as well as his breathing pathways, recreated from a CT image, and constructed an accurate model of the respiratory system. This way they could proof-check the sealing quality of the exceptional babies as well, and compare it to the sealing quality of the old masks. In that project about 400 children were photographed with a geometry-scanning device specially developed in Kimmel’s lab. The geometry scanning device or 3D camera is simple to operate and provided an accurate model in a fraction of a second. Intel licensed this 3D scanning technology form the lab and the resulting product was sold by the millions in the world under the name of RealSense. The images at the attached report demonstrate a 3D printed replica of the face geometry with a version of the mask in which there is a novel valve chamber for a pacifier and show a baby with a newly designed mask based on our method and using our geometry-sensing device.