Let me start with expressing my excitement about XCube – this is great leap forward and I highly appreciate the decision to make this publicly available. Thanks for all the work that went into this!
My question is with regards to input plugins to create XCubes. It seems that eo-learn and xcube are somewhat complementary in their scope. Are there plans to have input & export plugins for eo-learn’s EOPatch data format?
As I understand from skimming the open issues on the xcube github repo, there’s an issue translating non-uniform data layers in an EOPatch to an XCube. It would already help tremendously if only specific layers could be exported.
It would be helpful if you could provide your intended workflow, e.g. as pseudo code. Note that EOLearn already has a converter from xcube to EOPatch. I’ll find out whether this has been already released or is still an experimental feature. I let you know.
We currently leverage eo-learn’s SentinelHub integration to query, download, and store EO data. In a processing step we deploy an in-house water processor that heavily leans on in situ ground truth observations. After playing with XCube these past couple of days, I came to realize that with respect to the former tasks, eo-learn and xcube are fully redundant.
While eo-learn seems to emphasize the in-memory analysis of EO data, xcube is tailored towards efficient and scalable (cloud) computation, storage, and data delivery & visualization. Both projects are EU funded, open-sourced (MIT), under active development, and likely to be continued (not tied to specific project funding). Please correct me if you see any of this differently, I’m trying to navigate this as I go.
For me as a user it would help to be able to pick-and-choose specific feature from both frameworks, i.e. facilitated by seamless import/export for both data formats. However, at this stage it would take relatively little effort to migrate to XCube for the tasks at hand. Out-of-the-box WMS and viewer capabilities and built-in cloud capabilities may sway me in that direction.
In this context, I’d very much appreciate your view on how xcube and eo-learn compare currently, and, if you want to comment on this, how they might evolve.