3D data representation
Deep learning for 3D images

While images are classically represented as 2D matrices (possibly with vector values, as for color), various representations exist for 3D data: three-dimensional matrix, depth map, point cloud or mesh. Readers interested in point clouds may wish to refer to the article by Marcotegui and Serna [RE 187]. We won't deal with them here.

The three-dimensional matrix is a direct extension of the classic 2D image representation. Like a pixel, a voxel is an elementary volume in 3D space that contains certain information. The voxel grid is used for 3D images...