WineOCR

OCR-based tool, used to feed OpenWines open-data, using wine bottle labels images:

These scripts use nodejs, dependencies are managed via npm

Output:

$ node wineocr.js etiquette_3.jpg

CHÂTEAU MERCIER

CÔTES DE BOURG

APPELLATION CÔTES DE BOURG CONTRÔLÉE

2000

15H ml
ÇERÊRÏfiLL MIS EN BOUTEILLE AU CHATEAU

“l |
S.C.E.A. FAMILLE CHETY, PRODUCTEUR A SA|NT—TROJAN (GIRONDE) FRANCE  .

The OCR output is both displayed in the console and written in a newly created [IMAGE_PATH].ocr.txt file

Usage:

$ node wineocr.js [path] [layout_analysis_option] [language_code]

Example:

$ node wineocr.js ./examples/etiquette_3.jpg

which is the equivalent of the default options:

$ node wineocr.js examples/etiquette_3.jpg 3 fra

Argument details:

• both directory and single file path work, as first argument
• 3 is a layout analysis option for tesseract OCR
• fra is a language code. Available languages depend on your tesseract installation (see below).

Note that you can either process a whole directory:

$ node wineocr.js examples/

Installation

1/2 - Install tesseract

tesseract is an open-source project and a madatory dependency for WineOCR.

For Mac OS X:

brew install tesseract --all-languages

For other OS as Windows or GNU/Linux, and details about installing only certain languages packs, check out the tesseract-ocr Project website.

2/2 Install dependencies (node-tesseract, etc.)

npm install

Options at run:

layout_analysis_option argument (see psm argument in node-tesseract lib) tells Tesseract OCR binary to only run a subset of layout analysis and assume a certain form of image. The options are:

• 0 = Orientation and script detection (OSD) only.
• 1 = Automatic page segmentation with OSD.
• 2 = Automatic page segmentation, but no OSD, or OCR.
• 3 = Fully automatic page segmentation, but no OSD. (Default)
• 4 = Assume a single column of text of variable sizes.
• 5 = Assume a single uniform block of vertically aligned text.
• 6 = Assume a single uniform block of text.
• 7 = Treat the image as a single text line.
• 8 = Treat the image as a single word.
• 9 = Treat the image as a single word in a circle.
• 10 = Treat the image as a single character.

3 seems to be a good option for most wine bottle labels.

Considering tesseract alternative: OCR APIs

• Microsoft Project Oxford Visions OCR (w/ API)
• https://ocr.space
• https://cloud.google.com/vision/

License

MIT License - See license file.

Sources are available at OpenWines/WineOCR on Github.

Issues, support

Please check OpenWines/WineOCR/issues page on Github.