Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: meas_algorithms
-
Labels:None
-
Story Points:2
-
Epic Link:
-
Sprint:DRP S20-6 (May)
-
Team:Data Release Production
-
Urgent?:No
Description
There have recently been discussions regarding Defects where new defects are appended to the existing list of defects. This works fine but can result in a situation where lots of tiny bounding boxes could be represented by a few large bounding boxes. Many things become more efficient if the minimal set of bounding boxes is used by downstream processing.
This ticket will add a normalize method to Defects that will recalculate the defect bounding boxes. One implementation could be to generate a mask from the defects and then recalculate the defects from the mask. This would have the side effect of allowing direct comparison of two different Defects instances.
Attachments
Issue Links
Activity
Exactly. Internally you wouldn't need the readFits and then you would replace the internal list of bounding boxes. You do have to decide whether this is a mutable method or something that returns a new Defects object to the caller.
Comments left on PRs.
Expected defects in tests of cp_pipe had to be modified accordingly (due to normalization). See: DM-19903
Latest Jenkins: https://ci.lsst.codes/blue/organizations/jenkins/stack-os-matrix/detail/stack-os-matrix/31825/
There's a new version of falke8 that brings to light new errors that can only be found, for now, via Travis. After fixing a few of those, latest Jenkins: https://ci.lsst.codes/blue/organizations/jenkins/stack-os-matrix/detail/stack-os-matrix/31834/pipeline
Just reread it all - looks great 
Minor comment on one of the PRs.
From a discussion in the Monday calibration products call, this probably is just something like this out-of-class snippet:
myDefects = Defects.readFits(filenameToRead)
mi = afwImage.MaskedImageF(bboxMatchingDefects)
myDefects.maskPixels(mi, maskName="defectNorm")
myNewDefects = Defects.fromMask(mi, "defectNorm")