We present a novel deep-learning-based method for Multi-View Stereo. Our\nmethod estimates high resolution and highly precise depth maps iteratively, by\ntraversing the continuous space of feasible depth values at each pixel in a\nbinary decision fashion. The decision process leverages a deep-network\narchitecture: this computes a pixelwise binary mask that establishes whether\neach pixel actual depth is in front or behind its current iteration individual\ndepth hypothesis. Moreover, in order to handle occluded regions, at each\niteration the results from different source images are fused using pixelwise\nweights estimated by a second network. Thanks to the adopted binary decision\nstrategy, which permits an efficient exploration of the depth space, our method\ncan handle high resolution images without trading resolution and precision.\nThis sets it apart from most alternative learning-based Multi-View Stereo\nmethods, where the explicit discretization of the depth space requires the\nprocessing of large cost volumes. We compare our method with state-of-the-art\nMulti-View Stereo methods on the DTU, Tanks and Temples and the challenging\nETH3D benchmarks and show competitive results.\n