Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: afw, meas_base, pipe_tasks
-
Labels:None
-
Story Points:18
-
Epic Link:
-
Sprint:Science Pipelines DM-W16-1, Science Pipelines DM-W16-2
-
Team:Data Release Production
Description
We have an algorithm on the HSC fork that modifies AssembleCoaddTask to clip outliers in a much safer way, based on detecting contiguous regions in the difference between a non-clipped coadd and an aggressively-clipped coadd, and only rejecting pixels that are outliers in a single epoch.
One complication for this code transfer is that some of the coadd code has been refactored on the HSC side, and there may be code in hscPipe that duplicates much of what's in pipe_tasks. We may need help from Paul Price to resolve those inconsistencies before tackling this issue.
Once that's done, HSC code can be transferred from the following issues:
Attachments
Issue Links
- is triggered by
-
RFC-98 Rename or redefine assembleCoadd to make way for safe clipping
- Implemented
- relates to
-
DM-4391 Update testCoadds.py to accommodate changes in DM-2915
- Done
-
DM-4201 Documentation and technical debt in meas_base/PixelFlags.cc
- To Do
-
DM-14935 Thoroughly assess of the effects of turning on "junk suppression" at the visit level
- To Do
-
DM-4210 Create documentation and examples for SafeClipAssembleCoadd
- Done
-
DM-3703 Port suspect pixel flags to meas_base
- Invalid
-
DM-4209 Create unit tests for SafeClipAssembleCoadd
- Invalid
- mentioned in
-
Page Loading...
Nate Lust The revised code on meas_base is better, thanks. There are still a couple of issues which I think need attention: you should eliminate the duplicate `typedefs` and take another crack at the comment in the `fail()` function.
Beyond that, there are a lot of minor spacing and formatting issues. I feel almost bad about pulling you up on these trivial issues, but they do need fixing. I wonder if there are linters that could help you track down these issues in C++ code automatically – might be worth investigating. (There are certainly code formatters, like clang-format, but you want something that'll point out mistakes rather than attempt to reformat the whole file.)