Converting a PlateAcquisition takes very long - Githubissues

fmi-faim / faim-ipa

A collection of Image Processing and Analysis (IPA) functions used at the Facility for Advanced Imaging and Microscopy (FAIM)

BSD 3-Clause "New" or "Revised" License

9 stars 6 forks source link

Converting a PlateAcquisition takes very long #65

Closed tibuch closed 11 months ago

tibuch commented 11 months ago

Tested the MD and CV PlateAcquisition implementations on larger datasets and noticed that the implementations are slow. The time between creating an instance of a PlateAcquisition and the start of the stitching work is very long.

Pain points:

WellAcquisitions are not cached: https://github.com/fmi-faim/faim-hcs/blob/b9f1448260ac9c894dd4a2fe4829a440ef5ddba3/src/faim_hcs/hcs/cellvoyager/StackAcquisition.py#L72 but used at various points:
- https://github.com/fmi-faim/faim-hcs/blob/b9f1448260ac9c894dd4a2fe4829a440ef5ddba3/src/faim_hcs/hcs/acquisition.py#L63
- https://github.com/fmi-faim/faim-hcs/blob/b9f1448260ac9c894dd4a2fe4829a440ef5ddba3/src/faim_hcs/hcs/acquisition.py#L122 This is time-consuming, because building a WellAcquisition involves metadata parsing.
No caching of shapes and dtypes in WellAcquisition and PlateAcquisition.
Z-spacing computations runs on all files. One well, one channel and one site should be sufficient: https://github.com/fmi-faim/faim-hcs/blob/b9f1448260ac9c894dd4a2fe4829a440ef5ddba3/src/faim_hcs/hcs/imagexpress/StackAcquisition.py#L67