Changes to the deblender API

We are currently working on upgrading the deblender to process multiple bands at once, making it possible to use color information while deblending. This requires a few changes to the internal API and data products that are produced (but not persisted) in the deblender. Specifically, the PerFootprint and PerPeak classes are being replaced with the set of new classes to store deblender results. For users who run the deblending tasks (as opposed to calling lsst.meas.deblender.baseline.deblend) there will be no changes and the current deblender will continue to run as expected.

Once DM-8694 has passed review and is merged there will be an update on how to use the new internal deblender API for anyone interested, but for now it is important to see if the change in internal classes will affect anyone else’s research.

So if there is anyone whose work will be significantly impacted by changing PerFootprint and PerPeak please let us know so that we can take that into consideration. Note: with the current plan there are two objects very similar to PerFootprint and PerPeak, but they contain additional information needed for multi-color deblending and accessing them will be slightly different.

The branch with the changes is https://github.com/lsst/meas_deblender/tree/tickets/DM-8694

2 Likes

For those interested, ip_diffim also used the deblender internal components. So there were some minor changes in dipoleMeasurement.py that don’t affect the output but will require updating the stack for anyone developing on ip_diffim.