Use case/problem: when photographing exhibitions, I take pictures of the works + the museum labels (cartels, in French, nothing to do with drug trafficking). When I want to import the pictures of the works and create the linked Wikidata items, it’s tedious to read the info from the labels pictures and write them into Wikidata manually.
Idea: run an OCR tool on the folder to retrieve the info from the labels pictures and have them as ready to import as possible (through OpenRefine or QuickStatements or whatever mass import tool).
Current status: did some tests with Tesseract, some code is here