Pierre Moulon

I use the camera.transform data. I have re-oriented the arkit rotation matrix when I import the pose to OpenMVG as I explained to you. ARConfiguration.WorldAlignment was set to it's default...

Did you check that the camera poses you have provided are looking reasonable? You can use ConvertSfMDataFormat to do that `ConvertSfMDataFormat -i sfm_data.bin -o camera_trajectory.ply -E` You should check that...

Which version of ceres are you using: It seems that https://en.cppreference.com/w/cpp/algorithm/exclusive_scan is c++17 only. You can try set this line https://github.com/openMVG/openMVG/blob/develop/src/CMakeLists.txt#L129 to 17 rather than 11 and see how it...

=> https://github.com/ceres-solver/ceres-solver/issues/793

@KentaroUno Can you try to add this in multiview/CMakeLists.txt `target_compile_features(openMVG_multiview PUBLIC cxx_std_17)`

@KentaroUno Any feedback?

@AsDeadAsADodo Note that this dataset is most likely have GPS info that you can leverage to speed up the matching phase. See here for some tips: https://github.com/openMVG/openMVG/blob/develop/docs/sphinx/rst/software/Geodesy/geodesy.rst#usage-of-pose-prior

Can you tell us more about what do you call limited resources (which kind of platform (raspberry?))? Photogrammetry is having multiple bottlenecks and potential solution: - feature extraction -> use...

FYI, OpenMVS has been used on the phone by some application Here is one example https://play.google.com/store/apps/details?id=com.smartmobilevision.scann3d&hl=fr&gl=US Nowadays the best way to perform photogrammetry on mobile device is to mix results...

You either have to use CMake or link libraries and include folders by hand.