OpenMVG v2.0 "Rainbow Trout" is out

What's new:

• 21 PR has been handled.
• 32 issues/features enhancement has been completed.

OpenMVG core:

Logging

• Centralized LOGGER OPENMVG_LOG_INFO/WARNING/ERROR

Matching

• CBIR - Content Based Image Retrieval
  • VLAD global matching image indexing & retrieval enhancement [1,2,3]
  • Pre-Emptive feature matching [4]

Software

• One SFM binary to rules them All
  • access to Incremental, IncrementalV2 and Global from a single binary!
• Add utility to convert matches from .bin to .txt or .txt to .bin
• Improve OpenMVS interface
• ComputeStructureFromKnownPoses can now triangulate landmarks in sfm_data files (i.e computed by main_VO)
• Frustrum display
  • Top face colored
  • Add near/far bounds limits as command line parameters
• Dataset import
  • Add support of BlendedMVS/MVG dataset

Third-party

• Update Eigen to 3.4.0

Benchmarking

• Add an image collection matching benchmark binary

Thank you to @rjanvier, @rperrot,@cdcseacave, @nguyengiang41, @yocabon, @tjdahlke, @cenit, @TongLing916, @csparker247, @Cecilimon, @mitjap, @qingzhengzhuma, @kimoktm, @LWisteria, @flosincapite, @yurivict, @Hiltona99

[1]"Aggregating local descriptors into compact codes". H. Jegou, F. Perronnin, M. Douze, J. Sanchez, and P. Perez., C. Schmid. PAMI, 2012.
[2] "All About VLAD". R. Arandjelovic and A. Zisserman. CVPR 2013.
[3] "Revisiting the VLAD image representation". J. Delhumeau, P.H. Gosselin, H. Jégou, P. and Pérez. ACM Multimedia 2013.
[4] "Towards Linear-time Incremental Structure from Motion" C. Wu, 3DV 2013