https://github.com/damiencorpataux/foodopendata-digitization

Automated extraction of data from food labels pictures
https://github.com/damiencorpataux/foodopendata-digitization

Last synced: about 1 month ago
JSON representation

Automated extraction of data from food labels pictures

• Host: GitHub
• URL: https://github.com/damiencorpataux/foodopendata-digitization
• Owner: damiencorpataux
• Created: 2017-02-10T14:31:46.000Z (almost 8 years ago)
• Default Branch: master
• Last Pushed: 2017-02-10T15:48:53.000Z (almost 8 years ago)
• Last Synced: 2024-11-05T16:36:46.512Z (3 months ago)
• Size: 1000 Bytes
• Stars: 0
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# foodopendata-digitization

Automated extraction of data from food labels pictures

https://hack.opendata.ch/project/67

## Scope and process

- INPUT: food label picture

- PROCESS: identify the textual data areas from the picture

  - AI semantic segmentation problem here, ingredients and nutrition facts datasets exists for machine learning:

    - https://www.crowdai.org/challenges/openfood-nutrition-table-challenge#resources

    - https://www.crowdai.org/challenges/openfood-ingredients-list-challenge#resources

  - A blog post on machine learning to identify letters in an image (slightly off topic but the process is inspiring)

    - http://francescopochetti.com/text-recognition-natural-scenes/

- PROCESS: clean the cropped data area of the picture to make them OCRable (rotation, brightness/contrast, ...)

- PROCESS: OCR and parse the text stream

- OUTPUT: structured data, that should ideally be storable

## Legacy resources

- This guy made a prototype for OCRing text on an image with noise

  http://francescopochetti.com/text-recognition-natural-scenes/#second

- A ridiculously basic OCR prototype (might not be always accessible, this is a dev server)

  http://mien.ch:5555/fod/