Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-12697

Fix hollowed out cores of saturated stars in CompareWarp

    Details

      Description

      Thus far I've been clipping any any region for which we don't have information from the PSF-matched warp. This includes more area than direct warp because of matching kernel smears out the bad pixels and edges of calexps. Test allowing saturated pixels through. Merge if it works.

        Attachments

          Activity

          Hide
          yusra Yusra AlSayyad added a comment -

          There turned out to be two separate problems (both introduced by DM-12445):

          1) Holes in the template become holes in the resulting coadd. Solution: Reduce number of bad mask planes to default NO_DATA.

          2) Clipping ALL epochs. I originally thought this was impossible. I assumed if N=2 then neither epoch could be an outlier, but this assumption was wrong. Solution: reducing max number of epochs to a fraction of local N.

          Before:

          (Left = safeclip, Right = compareWarp after DM-12445)

          After:

          (Left = new compareWarp Image, Right = new compareWarp NImage)

          Show
          yusra Yusra AlSayyad added a comment - There turned out to be two separate problems (both introduced by DM-12445 ): 1) Holes in the template become holes in the resulting coadd. Solution: Reduce number of bad mask planes to default NO_DATA. 2) Clipping ALL epochs. I originally thought this was impossible. I assumed if N=2 then neither epoch could be an outlier, but this assumption was wrong. Solution: reducing max number of epochs to a fraction of local N. Before: (Left = safeclip, Right = compareWarp after DM-12445 ) After: (Left = new compareWarp Image, Right = new compareWarp NImage)
          Hide
          pschella Pim Schellart [X] (Inactive) added a comment -

          Looks good. As discussed IRL I think something that scales (slowly, and ideally with the probability of overlap) with N would be better in the long run.

          Show
          pschella Pim Schellart [X] (Inactive) added a comment - Looks good. As discussed IRL I think something that scales (slowly, and ideally with the probability of overlap) with N would be better in the long run.
          Hide
          yusra Yusra AlSayyad added a comment -
          Show
          yusra Yusra AlSayyad added a comment - I agree 100%. See https://jira.lsstcorp.org/browse/DM-12692
          Hide
          yusra Yusra AlSayyad added a comment -

          Thanks for the fast review. I've updated everything I plan on updating. pipe_tasks scons passes. Tested on a few problematic patches of the RC dataset.

          Summary of Jenkins runs:

          Show
          yusra Yusra AlSayyad added a comment - Thanks for the fast review. I've updated everything I plan on updating. pipe_tasks scons passes. Tested on a few problematic patches of the RC dataset. Summary of Jenkins runs: lsst_distrib: https://ci.lsst.codes/job/stack-os-matrix/27179/ passed ci_hsc: https://ci.lsst.codes/job/stack-os-matrix/27180 turns out I wasn't at tip ci_hsc: https://ci.lsst.codes/job/stack-os-matrix/27181 (currently running) lsst_distrib https://ci.lsst.codes/job/stack-os-matrix/27182 (currently running because I changed some things)
          Hide
          yusra Yusra AlSayyad added a comment -

          RC tracts 9813 and 8766 are available in: /datasets/hsc/repo/rerun/private/yusra/psfMatching/DM-12697/

          Show
          yusra Yusra AlSayyad added a comment - RC tracts 9813 and 8766 are available in: /datasets/hsc/repo/rerun/private/yusra/psfMatching/ DM-12697 /

            People

            • Assignee:
              yusra Yusra AlSayyad
              Reporter:
              yusra Yusra AlSayyad
              Reviewers:
              Pim Schellart [X] (Inactive)
              Watchers:
              Paul Price, Pim Schellart [X] (Inactive), Yusra AlSayyad
            • Votes:
              0 Vote for this issue
              Watchers:
              3 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved:

                Summary Panel