Need help accessing an Elsevier journal article?

Click here

Image Recognition + Label This

If the 5,000 labels in the Professor Amerine wine label collection is any indication, it can take a lot of bottles to become an educated wine consumer. And unless you’re tasting wines like Professor Amerine (or our V&E students!), there’s a good chance that when you walk down the wine aisle in the grocery store, decision fatigue sets in. In the App Store, Vivino and Delectable are two apps hoping to make that moment a little less painful. Simply take a photo of the label, and they’ll return reviews and ratings. And wine geeks everywhere rejoice! Pretty cool.

So one of the questions that we’ve received from Label This transcribers is whether the technology that’s powering those apps might work for us. Unfortunately, the answer is a big no. Both Vivino and Delectable depend on an image database for pattern-recognition. So when someone takes a photo of a label, the site is mapping it to existing labels in the database.

So while these sites are using OCR to help the database ‘read’ the text on the label, the real juice is coming from that database. (Get the story on how Vivino built their database from co-founder Theis Søndergaard on Medium)  For our labels, because of their uniqueness, there isn’t any existing database to pull from and since there is so much variance in the type and illustrations on the labels, we really needed human hands (and human brains!) to tell us what is on the label. So what are you waiting for? We need your help to Label This!