Chapter 12: Stereo correspondence

Open-source software for the latest stereo matching, multi-view, and monocular depth inference algorithms usually accompanies recently published papers. Lists of the most recent and best performing algorithms can be found on the leaderboards associated with the most popular benchmarks such as Middlebury, KITTI, ETH3D, and Tanks and Temples, which are discussed in Appendix C.1 and Tables 12.1 and C.1. algorithm

C.2 Software

Both MVE (https://www.gcc.tu-darmstadt.de/home/proj/mve) (Fuhrmann, Langguth et al. 2015) and COLMAP (https://github.com/colmap/colmap) (Schönberger, Zheng et al. 2016) provide complete 3D reconstruction pipelines that include structure from motion, multi-view stereo densification, mesh generation, and texturing. A review of earlier packages can be found in Furukawa and Hernández (2015).

• A number of high-quality commercial photogrammetry packages such as CapturingReality, ContextCapture, Metashape, and Pix4D, which grew out of computer vision research labs, provide similar functionality. $ ^{7} $