Bottle Reader

[dcramer]

We'd like to be able to take a photo of a label of a bottle and use it to create an improved UX. Inspiration comes from Vivino and Untappd.

Ideally it would let us:

• Identify the bottle for search
• Extract chunks to prepulate the Add Bottle flow
• Flatten and capture an image to store w/ the bottle

Some prior art:

• https://github.com/AntoninLeroy/wine_label_reader_toolkit
• https://github.com/OpenWines/OpenWinesOCR/blob/master/wineocr.js
• https://hackernoon.com/how-to-unwrap-wine-labels-programmatically-31c8c62b30ce
• https://services.tineye.com/WineEngine
• https://towardsdatascience.com/how-to-read-a-label-on-a-wine-bottle-using-computer-vision-part-1-25447f97a761
• https://aws.amazon.com/textract/

[dcramer]

Tested out WineEngine to see if it would let us bootstrap real quick, but its probably overly biased around wine.

It did extract labels on two examples, but did very little OCR (and I dont see any APIs that would return more text):

[dcramer]

This is likely going to just be in a mobile-native app as its prohibitively complicated to do via web JS.