You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 15 Next »

Data access example

from psana import DataSource
ds = DataSource(exp='tmoc00318',run=10, dir='/cds/data/psdm/prj/public01/xtc')
orun = next(ds.runs())
det = orun.Detector('epix100')
for evt orun.events():
    raw = det.raw.raw()

Direct detector interface methods

Hidden for AMI mask methods

For implementation of AMI most of the area detector interface methods are hidden - name starts with underscore "_". In analysis code these methods can be used directly:

DTYPE_MASK = np.uint8
o = det.raw
mask = o._mask_default(dtype=DTYPE_MASK)
mask = o._mask_calib()
mask = o._mask_calib_or_default(dtype=DTYPE_MASK)
mask = o._mask_from_status(status_bits=0xffff, gain_range_inds=(0,1,2,3,4), dtype=DTYPE_MASK, **kwa) # gain_range_inds for appropriate detectors
mask = o._mask_neighbors(mask, rad=9, ptrn='r')
mask = o._mask_edges(width=0, edge_rows=1, edge_cols=1, dtype=DTYPE_MASK, **kwa)
mask = o._mask_center(wcenter=0, center_rows=1, center_cols=1, dtype=DTYPE_MASK, **kwa)
mask = o._mask(status=True, status_bits=0xffff, gain_range_inds=(0,1,2,3,4),\
               neighbors=False, rad=3, ptrn='r',\
               edges=True, width=0, edge_rows=10, edge_cols=5,\
               center=True, wcenter=0, center_rows=5, center_cols=3,\
               calib=False,\
               umask=None,\
               force_update=False, dtype=DTYPE_MASK)

Usage of mask in methods det.calib and det.image

kwa = {'status':True, 'status_bits':0xffff, 'gain_range_inds':(0,1,2,3,4),\
       'neighbors=False, 'rad':3, 'ptrn':'r',\
       'edges':True, 'width':0, 'edge_rows':10, 'edge_cols':5,\
       'center':True, 'wcenter':0, 'center_rows':5, 'center_cols':3,\
       'calib':False,\
       'umask':None,\
       'force_update':False}
o =  = det.raw
a = o.calib(evt, cmpars=(7,2,100,10), **kwa)
a = o.image(evt, nda=None, **kwa)

Cumulative mask example

epix10ka_raw_calib_image mask -e ueddaq02 -d epixquad -r211 -o fig-ueddaq02-r211-epixquad-mask.png -P

umask = np.ones((4, 352, 384), dtype=np.uint8)
umask[3,100:120,160:200] = 0

mask = det.raw._mask(status=True, status_bits=0xffff, gain_range_inds=(0,1,2,3,4),\
                     neighbors=True, rad=5, ptrn='r',\
                     edges=True, edge_rows=10, edge_cols=5,\
                     center=True, center_rows=5, center_cols=3,\
                     calib=False,\
                     umask=umask,\
                     force_update=False, dtype=DTYPE_MASK)
mask += 1 # for visibility of the mask 0 and 1 relative to image background

Class Mask

Alternative usage of mask methods is implemented in class Mask. Class Mask is a wrapper of hidden for AMI mask methods, which can be used to set dedicated mask parameters separately from all other detector interface optional parameters.

Instantiation of the Mask object

Object of the class Class Mask is instantiated for requested list of optional parameters,  evaluates and caches combined mask in det._mask_ property. If this property is None - mask is re-evaluated. 

from psana.detector.mask import Mask, DTYPE_MASK

m = Mask(det,\
         status=True, status_bits=0xffff, gain_range_inds=(0,1,2,3,4),\
         neighbors=True, rad=5, ptrn='r',\
         edges=True, width=0, edge_rows=10, edge_cols=5,\
         center=True, wcenter=0, center_rows=5, center_cols=3,\
         calib=True,\
         umask=test_umask(det),\
         force_update=False, dtype=DTYPE_MASK)

m = Mask(det) # minimal version.

Class Mask access methods

All other methods of the class Mask are shotcuts to det._mask* methods without "_".

m = Mask(det)
m.set_mask(**kwa) # forced update of cached mask, **kwa - the same list of keyword arguments as in class Mask w/o det.
mask = m.mask(**kwa) # returns cached mask, **kwa - the same list of keyword arguments as in class Mask w/o det.
mask = m.mask_default()
mask = m.mask_calib_or_default() # if available returns mask from "pixel_mask" calibration type, otherwise array of ones. 
mask = m.mask_from_status(status_bits=0xffff, gain_range_inds=(0,1,2,3,4), dtype=DTYPE_MASK)
mask = m.mask_edges(width=0, edge_rows=1, edge_cols=1, dtype=DTYPE_MASK)
mask = m.mask_center(wcenter=0, center_rows=1, center_cols=1, dtype=DTYPE_MASK)
mask = m.mask_neighbors(mask, rad=9, ptrn='r')

A few test examples can be found in psana/detector/testman/test_mask.py

Implementation notes

  • Default combined mask parameters (status=True, neighbors=False, edges=False, center=False, calib=False, umask=None) are set to define mask for pixel_status only.
  • Mask defined as m = Mask(det) and property det._mask_ is cached for combined/cumulative mask retrieved by the method mask=m.mask().
  • Method m.set_mask(**kwa) or its equivalent mask = m.mask(force_update=True, ...) can be used to update cached mask, if necessary.
  • Method m.mask(...) applies mask_neighbors after mask_from_status only. It this is not enough, mask needs to be constructed using separate m.mask_*(...) methods.
  • Most of methods mask = m.mask_*(...) do not cache intermediate results and consume time for specific mask evaluation. To optimize computing time in analysis it would be wise to evaluate mask once per run only.
  • Module psana.detector.UtilsMask contains useful methods for manual mask building, for example merge_masks(...).
  • By default mask is used in common mode correction, but it is not applied to data in det.raw.calib(...)

References

  • No labels