Skip to content

Feat: Upgrate agipd corr

Mikhail Karnevskiy requested to merge feat/new_agipd_corr into master

@kluyvert @ahmedk

New correction algorithm uses Multiprocessing with shared memory.

N processes for file manipulations (reading and writing) and N*M processes for correction (M processes for each file) are started at the beginning of the correction process. Therefore N files are processed in parallel, while all their images are in the shared memory.

This MR requires additional functionality of pyDetLib, which is available in branch https://git.xfel.eu/gitlab/karaboDevices/pyDetLib/tree/feat/add_cython_functions.

Main changes:

  • Parallelization (using multiprocessing module) over files and chunk of images
  • Load all images from file in RAM. Use shared memory (sharedmem module).
  • Add common-mode correction
  • Add flags for all corrections
  • Move statistics and preview part from agipdlib.py to notebook. Show first train in data.
Edited by Mikhail Karnevskiy

Merge request reports