I have to confess that my crowd-sourcing experiment was an abject failure. Not a flicker of assistance from my so-called friends in the reading group :-( Of course they are probably more wisely reading the book instead. Rephrasing the old Harvard Graffiti I guess I'd rather make websites for books to read than read books.)
But I pressed on obsessively to get the illustrated work completed. So far nearly all the Artists and Locations have been linked and about 95% of the illustrations are linked. I've linked to WikiMedia Commons images where possible, and to Wikpedia pages as derived links, with overriding of the page name in a few cases. The names of Artists and Locations have been normalised to support this link, a task which has also improved the data quality, removing some errors introduced by the OCR and normalising names from Berger's book.
These names are also used to create links to dbpedia, from where data on birth date and place,death date and place,location ec is often available. I'm unsure whether to derive data from dbpedia via runtime SPARQL query or to use it for batch update - I think the latter since response time to SPARQL queries is variable and data is incomplete.
I've also added subject tags to the paintings, based again on wikipedia pages. Subjects change the ER model because they are not instantiated in the database, being only concatenated values in a 'subjects' string.
In the process I've come across a bewildering number of old master sites. Many of these are commercial sites marketing hand-painted reproductions.