Closed vigji closed 8 months ago
Thanks for the suggestion!
Right now the software is relatively format-agnostic. Most of the io functions ingest numpy arrays directly, rather than paths to data in a specific format. So in a sense it is no more format-specific than matplotlib.
I agree that interfacing with standard formats would be useful through. To that end, we developed a conversion tool that generates SNUB datasets directly from NWB files (see https://github.com/calebweinreb/SNUB/pull/17). See documentation here: https://github.com/calebweinreb/SNUB/blob/dev/docs/source/nwb.rst
I see the point that you make in referring to it as an almost matplotlib-like tool in being data-agnostic. Still, I feel that including pipelines to generate the data from nwb files does lower the energy barrier for adoption for new users, and gives a good tool to dig into open datasets. I feel that this is addressed now!
As the package aims at helping out people dealing with their own recordings and experiments, it would greatly benefit from supporting standard neuro data formats.
I believe that the impact of the tool would be way larger if there were some more integration with the existing ecosystem of data formats, in the current shape it feels a bit a lab-specific tool (format-wise), instead of forcing the user to structure their dataset to the highly specific requirements of the package. I could recommend for example look for at least proof-of-principle integration with NWB, or SpikeInterface.
A good way of smoothing this up would be to showcase the software on openly accessible datasets, instead of the custom datasets provided as examples.