Openmvg: OpenMVG Pipe Line & Reconstruction using image sequence

I Tyler, OpenMVG is computer vision library that focus on strong implementation of multiple view geometry algorithms.

It provides:

• 2 Structure from Motion pipeline:
  
  
  
  • an Incremental Structure from Motion chain [ACSfM]_ (ACCV 2012),
  
  
  • a Global Structure from Motion chain [GlobalACSfM]_ (ICCV 2013).
  
  
• tools to visualize:
  
  
  
  • features,
  
  
  • photometric/geometric matches correspondences,
  
  
  • features tracks.
  
  
• export to existing Multiple View Stereo-vision pipeline:
  
  
  
  • [PMVS]_, CMPMVS.
  
  
• tools to build your own SfM pipeline:
  
  
  
  • geometric solvers, robust estimators ...
  
  

As you can see here: https://github.com/openMVG/openMVG/blob/LInfinityCV/docs/sphinx/rst/software/SfM/SfM.rst
OpenMVG can find the a 3D point cloud (structure) and the motion of a series of pictures.
According some users, openMVG SfM chain have been used successfully with up to 1600 pictures.

To infer the 3D model you have after to perform some post-processing, known as MVS Multiple View Stereo (PMVS, CMPMVS). OpenMVG provide ready to use compatible export for those binaries.

At the end you will be able to produce model like the one I have done here: https://sketchfab.com/show/2cde608f89424a7786bdc751857f8dfd

Structure from Motion openMVG chain focus on accuracy and not on speed, even they can be fast.

So contrary to the following alternatives:

• VisualSfM, Changchang Wu, "Towards Linear-time Incremental Structure from Motion", 3DV 2013
• Bundler, Noah Snavely, Steven M. Seitz, Richard Szeliski. Modeling the World from Internet Photo Collections. IJCV, 2007.
  OpenMVG want provide an easy to understand, launch and modify source code.

Hoping the answer is complete.

@pmoulon Thank you for your detailed answer. I'm a tiro in this field so I want to ask if there's any 3d reconstruction sample (C++) using OpenMVG as a reference?

@fansays OpenMVG is made in C++.
There is some Python scripts that just allow to launch more easily the command line tools.
It's more easy to use first the provided python scripts for which you just have to specify an input image dir and an output dir.
Once you master that you can play and have more control on the pipeline by using directly the binaries.