Language-independent format for dataset

[jzaremba]

This would facilitate interfacing with other projects and applications (ImageJ for example) and could potentially allow datasets to be read without needing a full SIMA install.

Store all metadata (everything currently in dataset.pkl, most of sequences.pkl) in a single file (XML, YAML, JSON?) and any actual data (displacements, rois) in a separate file or files (h5, csv?).

Alternatively, consider other Open data formats, potentially further integrated with experiment metadata.

[pkaifosh]

This is a good idea. It will also make it easier to inpect/edit datasets by hand. Do you want to start trying this out on a new branch? JSON?

[jzaremba]

I can look in to it a bit, but it probably won't be a priority for me right now. Let's make sure the format we pick will work with any of our potential future applications.

ImageJ?
Thunder?
Ilastik?
some other 3D volume viewer?