Open piranna opened 7 years ago
I think it needs some discussion, but it's a good idea.
I guess the simplest way would be to maybe make a directory called ~/.whereami/areas, in which we have an area name (living.txt) with the names of the subregions (couch1, couch2) in it rather than sample data?
I would certainly welcome a PR.
You could just as easily just do this all from a label standpoint. For example, name it living/couch1
and living/couch2
- this would automatically tell you where you are as well as where that is as well.
But this is a fascinating project :)
For example, you can be able to test not only on different couchs on the same room, but also define them on the same room so the train data apply to the bigger region too (couch1 + living, couch2 + living). It could be possible to return the different "closures" from more specific to more generic, or viceversa.