Tomaz-Vieira
commented
4 years ago I have a feeling that we should move away from 'hosting' the datasets and just allowing the user to enter the URLs of the datasets he's interested in. This would make life a bit harder on us since it wouldn't be trivial to figure out if some URL can be processed by some .ilp project, but maybe we can leave such responsibility on the hands of the user. Alternatively, we could have some DataSource implementations also available on cloud-ilastik, so that it can 'connect' to the user's URL, check dimensionalities and fail faster in case of bad data
cloud_ilastik currently features three public datasets from the knowledge graph:
the source should be obvious when viewing datasets.