Spreadsheet/table with inputs/outputs that go into every existing (or planned) metric in BTK
A refactored metrics module where every metric can be used independently of each other and where input/output are clear to the user
(If time) New set of metrics that also capture information about unmatched objects
Resources and skills needed
Anyone with an interest in brainstorming and designing new blending metrics is welcome to join!
Python experience is helpful for implementing your own metric into BTK
BTK experience is not necessary but might be helpful
Detailed description
We want to make the metrics module in BTK more useful for the wider blending community. Currently, the functions in the metrics module are quite hard to use without running the entire BTK pipeline. This is undesirable when people want to evaluate the results of their deblending algorithm that they might have produced using different sets of catalogs or images than the ones inside BTK.
There are two main steps to accomplish this (1) Clarify the input and output of every metric we want implemented in BTK (existing or planned ones) (2) Brainstorm a way to organize the metrics module so that it's more accessible. We will spend the chunk of the time in thinking and re-designing the metrics module. If time allows, we can also design new metrics that capture other important errors in deblending beyond what is already implemented in BTK e.g. ambiguously blended objects.
Design new metrics and enhance the metrics module in BlendingToolKit
We will focus on enhancing the metrics module of BTK and designing new blending metrics.
Contacts: @ismael-mendoza Time: All day Main communication channel: #blending_toolkit GitHub repo: https://github.com/LSSTDESC/BlendingToolKit
Goals and deliverable
Resources and skills needed
Detailed description
We want to make the metrics module in BTK more useful for the wider blending community. Currently, the functions in the metrics module are quite hard to use without running the entire BTK pipeline. This is undesirable when people want to evaluate the results of their deblending algorithm that they might have produced using different sets of catalogs or images than the ones inside BTK.
There are two main steps to accomplish this (1) Clarify the input and output of every metric we want implemented in BTK (existing or planned ones) (2) Brainstorm a way to organize the metrics module so that it's more accessible. We will spend the chunk of the time in thinking and re-designing the metrics module. If time allows, we can also design new metrics that capture other important errors in deblending beyond what is already implemented in BTK e.g. ambiguously blended objects.