The current state of scarlet and looking toward the future

Abstract

Science pipelines made meas_extensions_scarlet the default deblender for multi-band deblending in weekly w_2021_05, which is powered by the scarlet package. scarlet was co-created by science pipelines and Peter Melchior’s group at Princeton University with the dual goal of deblending multi-band images in HSC/Rubin Observatory data and future joint processing with a fusion of observations from both ground and space-based telescopes as well as other instruments like IFUs and grisms. As scarlet has become an increasingly more powerful tool, the chain of classes and commands required to build the objective function has become substantially more complex in order to properly handle the wide variety of applications that scarlet supports (both now and in the near future). This technote describes the current complexity of scarlet models, the subset of those models actually used in the science pipelines, and a proposal for using a more simple model in meas_extensions_scarlet to potentially speed up development, runtime, and improve the memory footprint.