Hello everyone,
I would like to report my successful testing of running RSGISlib on the new "Linux inside Windows" feature of Windows 10. If you haven't heard about it, it was introduced on the Anniversary Update:
"This isn’t a virtual machine, a container, or Linux software compiled for Windows (like Cygwin). Instead, Windows 10 gains a Windows Subsystem for Linux, which is based on Microsoft’s abandoned Project Astoria work for running Android apps on Windows. Think of it as the opposite of Wine. While Wine allows you to run Windows applications directly on Linux, the Windows Subsystem for Linux allows you to run Linux applications directly on Windows."
The Windows disks can be accessed on bash at /mnt/(drive letter)/:
I followed the instructions above to enable bash, and then installed anaconda and RSGISlib as I would on a Linux machine. I did ran into some of the dependency problems people have been encountering lately, but was able to fix them by installing everything at once from condaforge and au-eoed, and then running "conda update -c au-eoed -c conda-forge --all" while my environment was activated.
The only thing that won't work is Tuiview, since there is no graphical environment available. But as you can easily work on Windows folders, QGIS or other software can be used for visualization, and it seems there is a Tuiview binary for Windows.
To test the installation, I ran all scripts from the RSGISlib training course given by Pete during the 23rd ALOS K&C meeting, last January in Tokyo:
01_BandMaths.py
02C_PALSARStack.py (it did use all available cores!)
03B_MosaicImages.py
04_Resample2Landsat.py
05_ValidImageryMask.py
06A_RasterToVector.py
06B_VectorToRaster.py
07_Subset2ROI.py
08_MaskImage.py
09_BandSubset.py
10_LeeFilter.py
11_Convert2dBs.py
12_CalcReflIndices.py
13_CombinedPreProcessing.py
14_PixelBasedClass.py
15_SegData.py
16_PerformObjClass.py
All scripts ran flawlessly, and with good performance (under a minute each, on a 6-core i7 5820k, 64GB RAM).
This is pretty exciting, as it seems we can now use RSGISlib on any windows machine without the hassle of dual booting or vm-ing, no noticeable loss of performance.
Pete, Daniel, let me know if you would like me to do any specific tests.
Cheers,
--Thiago Sanna Freire SilvaProfessor Assistente Doutor
Instituto de Geociências e Ciência Exatas (IGCE)
Universidade Estadual Paulista (UNESP)
Rio Claro, SP - Brasil