colmap

Why we included this project

COLMAP has been the standard way to turn plain photographs into 3D geometry for years, and much of computer vision that needs camera poses leans on it. Give it a set of overlapping images, ordered or unordered, and it works out where each photo was taken and rebuilds the scene as a sparse point cloud, with an optional dense model on top. The GUI, command-line tool, and pycolmap Python bindings cover everything from a one-off lab reconstruction to an automated pipeline feeding NeRF or Gaussian Splatting training. It is genuinely general-purpose: drone photos, tourist snapshots, and carefully staged capture sets all come out with usable poses and geometry. Good documentation and sample datasets keep the first attempt from turning into a slog.