search

poldracklab / pydeface

defacing utility for MRI images
MIT License
112 stars 42 forks source link

Detail the list of FSL dependencies #44

Closed mateuszpawlik closed 2 years ago

mateuszpawlik commented 2 years ago

I want to use pydeface in a bigger context of data acquisition. I've added FSL into my Dockerfile. The problem is that FSL takes ~2.5h to install and ~16GB of space. I've posted some of the issues in https://github.com/ReproNim/neurodocker/issues/450

I'm wandering if I could at least reduce the container size by specifying more precise FSL dependencies, for example, FLIRT.

Here is the list of top-level directories in FSL directory and their sizes:

root@6778f59b8716:/opt/fsl-6.0.5.1# du -sh *
4.0K    LICENCE
4.0K    README
4.4G    bin
8.0K    build
3.1M    build.log
152K    config
2.3G    data
188M    doc
5.0M    etc
339M    extras
6.3G    fslpython
4.7M    include
1.6G    lib
264K    python
324K    refdoc
257M    src
2.4M    tcl
kaczmarj commented 2 years ago

@mateuszpawlik - neurodocker has a minify function designed to minimize docker images for particular commands. please see https://www.repronim.org/neurodocker/user_guide/minify.html for an example.

mateuszpawlik commented 2 years ago

I've successfully minified a Docker image for pydeface and FSL 6.0.5.1 (my image has other software included). It has 2GB now. Thank you @kaczmarj for the help. It took quite some time though. If any of this could be useful please let me know what could I do. Other than that we can close this issue.