Systems and methods for generating a depthmap for images of structures in a geographic area are provided. In some aspects, a method includes dividing the area into cells. Each of the cells is visible in no more than a threshold number of images. The method also includes: generating, using multi-view stereo processing, a set of three-dimensional (3D) points of structures in each of the cells based on the images associated with a corresponding one of the cells; generating a mesh model of the structures in each of the cells based on the set of 3D points for a corresponding one of the cells; identifying one or more of the mesh models visible in a corresponding one of the images; and generating a depthmap for each of the images based on the identified one or more mesh models visible in a corresponding one of the images.