A system and method is provided for generating textured 3D building models from ground-level and orthogonal imagery. Ground-level images for the sides of building objects are collected to form a 3D dense point cloud for identification of key architectural features, corresponding key façade geometry planes, and generation of a 3D building façade geometry. Orthogonal images are correlated to the 3D building model providing a properly geo-positioned, scaled and textured 3D building model.