COLMAP is an automatic reconstruction tool that takes a folder of input images and produces a sparse and dense reconstruction in a workspace folder

Given a set of overlapping photos or video frames of the same scene it estimates:

  1. Camera positions - where each photo was taken from
  2. Camera intrinsics - focal length, distortion, and related parameters
  3. Sparse 3D point cloud - rough 3D structure from matched image features

COLMAP is used to turn images into a 3D scene representation