In order to (potentially) use Google Cloud Vision as an image labeling provider, we need a concept mapping from its internal entity IDs to Wikidata IDs.
The GCV labeling API returns an AnnotateImageResponse with each entity represented by an EntityAnnotation object. According to the spec, entity IDs (mids) are opaque, but in practice these are Freebase IDs. (That said, there's technically no guarantee that they will remain stable.)
A Freebase->Wikidata mapping from 2013 is provided by Google. We need to create a script to identify any missing Wikidata IDs in the mapping and investigate what's happened to them. Once we're confident in the mapping, we can store the data in a MySQL table for use in MediaWiki.