Stand out from other ML libraries

[miquelmassot]

Just to make people aware, there are other python ML libraries that could be considered as competitors and would be interesting to write down what makes scivision special and stand out from them in the documentation - so that the end-user understands where scivision stands.
The libraries I could think of are:

• PyCaret
• H2O AutoML
• Plain pyTorch / Coffee / Keras

The arguments to use could be

• Learning curve
• Ease of deployment
• Speed

[edwardchalstrey1]

A point that is relevant to this and #168

Posting here what I initially wrote on Turing internal channel because this is relevant to any people interested in scivision:

one thing that it would be nice to have from scivision phase 2 would be a clearly defined, easy to explain USP (or whatever the open source software equivalent of a unique selling proposition is). The suggestion that Scott made in the meeting was that that Evie’s idea to use Mapreader for plants was only possible to have come up with because of the Turing network. This has me thinking that scivision, once a mature project/community, could effectively be providing the role that the Turing institute did in that case, sparking novel cross-disciplinary links in computer vision associated fields. Perhaps I’m stating the obvious there! It might however shift the emphasis towards building the academic community involvement and getting lots of things submitted to the catalog, as oppose to building a swiss army knife python package for installing/using/submitting catalog models.