• The reconstruction initialization is done thanks to an abstract interface, SfMSceneInitializer. The reconstruction can now start from 2 to n images (n-uplets or existing poses).
• The triangulation stage considers the entire scene tracks.
• The resection stage is based on 2d-3D matching confidence.
• The reconstruction can start with existing camera poses.
• This new engine is easier to read and to customize (compact and use more abstraction).

software/localization:

• Extend the localization software to support any camera type in the command line #1210

software/SfM:

• Add openMVG_main_SfMInit_ImageListingFromKnownPoses to init OpenMVG sfm_data scene (views, intrinsics & poses) from GT dataset #1195:
  • DTUMVS, ETH3D, Kitty, Middlebury, Strecha
• Remove useless code in IncrementalSfM #1254
• Add functions to export tracking data to sfm_data landmarks. enhancement #1292
• Move export2X and exportX binaries to software/SfM/export directory enhancement #1268
• Move the matcher out of a loop for to be more efficient in main_ListMatchingPairs.cpp #1172

Build enhancement

• User modern CMake recommendations #1176 #1146 #1235 #1147 #1221 #1269
• Fix shared libraries build #1339 #1293 #1286 #1358
• Enhance build documentation to support external libs build by Microsoft/VCPKG #1327
• Support Ceres-Solver 2.0 API changes #1323
• Enhance support of external JPEG, PNG, and TIFF libraries #1266

Samples

• Rewrite main_pano_converter.cpp to use the Spherical camera model enhancement #1314

Continuous Integration

• Enhance Travis XCode build #1170

Documentation

• Improve build and geodesy documentation #1337 #1346 #1251

UI

• Fix some minor issues #1224