MR external photogrammetry
I started to draft some nodes to utilize the other libraries: https://github.com/natowi/meshroom_external_plugins/tree/master/draft It is far from complete, but it is something I work on from time to time. It is possible to use Meshroom as the frontend for other photogrammetry pipelines, writing the required nodes is not too difficult. The only tricky part is to handle the workdirectory vs the meshroom one node one folder structure and making sure the output format is compatible with the rest of the Meshroom pipeline.
Converters not implemented in Alicevision but available in openMVG:
openMVG_main_openMVG2Colmap
openMVG_main_openMVG2openMVS
https://openmvg.readthedocs.io/en/latest/software/MVS/MVE/#export-to-mvs-texturing
requires no cuda but is slower
TODO
mve denserecon -> https://github.com/natowi/meshroom_external_plugins/wiki/MVE2EXR -> MR Meshing
Windows binaries https://www.gcc.tu-darmstadt.de/media/gcc/code/mve-20160517-win64.zip
Shading-aware Multi-view Stereo does not require cuda
https://github.com/flanggut/smvs https://github.com/flanggut/smvs/issues/2 https://github.com/flanggut/smvs/files/1473536/smvsrecon.zip "This project is intended as an alternative multi-view stereo step and replaces the dmrecon and scene2pset applications with the smvsrecon"
https://peterfalkingham.com/2016/10/29/photogrammetry-testing-7-smvs-mve/
Dense Reconstruciton without CUDA requirement. Algorithm is somewhat outdated (released in 2010/11)
http://francemapping.free.fr/Portfolio/Prog3D/PMVS2.html (without cuda)
https://www.di.ens.fr/pmvs/documentation.html
https://colmap.github.io/tutorial.html#dense-reconstruction
https://github.com/pmoulon/CMVS-PMVS/tree/master/binariesWin-Linux
https://github.com/qedsoftware/pmvs-gpu (with cuda support, tested with ubuntu only)
Note: I could not get this to work.
^^
OpenSfM is a Structure from Motion library written in Python (no cuda) https://github.com/mapillary/OpenSfM https://www.opensfm.org/docs/using.html https://www.opensfm.org/docs/building.html
https://github.com/mavmap/mavmap
"Using a combination of points and lines has the potential to improve SfM for indoor-and urban environments, where distinctive feature-points are rare" https://pure.tugraz.at/ws/portalfiles/portal/2694275/hofer_gcpr15.pdf