Reusable Workflows

[Jaapel]

Hi @scottstanie !

I previously used isce2, which also had this Workflow part, where community members could create their own workflows based on the framework that isce2 offered some kind of generic workflow component, where you could compose your own interferometric workflow. isce3 seems more like a standalone package of algorithms, am I correct? If so, I like that isce3 keeps its scope smaller.

What are plans for people sharing their workflows using isce3 algorithms? Sweets is of course an example of a workflow.

For example, I would like to be able to run a goldstein filter, and add other options that dolphin exposes. But I may run into other requirements that perhaps require adding additional steps in the sweets workflow. What do you see as the scope for sweets? As the GitHub description says:

Workflow for creating unwrapped interferograms from Sentinel-1 geocoded SLCs.

Which can of course be implemented in many ways :)

[scottstanie]

Hi @Jaapel , sorry for the long delay in response:

I believe right now, the plan for isce3 workflows is to have the nisar.workflow subpackage be a set of production-ready examples: https://github.com/isce-framework/isce3/tree/develop/python/packages/nisar/workflows

For some background: there may be a split of the "nisar" package from the "isce3" core library, but it has not fit into the NISAR development schedule (which is, so far, the only source of funding for isce developement). I don't think there is a plan to make some top-down, Abstract Base Class-based "Workflow" with isce3. The NISAR algorithm team already has their hands full producing the planned data products, and making "a cleaner version of the isce2 contrib folder" seems way out of scope.

I do think Sweets could be a little better refactored so that you could have your new options that you want; I know that I'm already seeing some differences in how I want to run it now vs. 2 years ago when I put it together.

I think the biggest immediate change for me is wanting different input sources:

• I made this when the ASF Burst service was being created, so I would now rather use that instead of the full SAFE download option
• Also, I'd want to be able to use the pre-made OPERA geocoded SLCs if I'm processing an area in north america. That would cut the workflow by about half