Nutrition facts table data extraction/Hackathon: Difference between revisions

From Open Food Facts wiki
No edit summary
No edit summary
Β 
Line 3: Line 3:
== Introduction ==
== Introduction ==


A team of Microsofties is working on using Microsoft Form Recognizer to extract data from nutrition facts table pictures to pre-populate [https://world.openfoodfacts.org/hunger-game?type=nutrition Open Food Facts' Hunger Game data entry and validation form].
A team of Microsofties were working during summer 2020 on training custom model to extract data from nutrition facts table pictures to pre-populate [https://world.openfoodfacts.org/hunger-game?type=nutrition Open Food Facts' Hunger Game data entry and validation form].


== Prototype ==
The results were not satisfactory, and the model was not integrated production.


* https://world.openfoodfacts.org/hunger-game?type=nutrition
'''More recent work on nutrition data extraction is done on the on-going [[NutriSight]] project.'''
Β 
== Code ==
Β 
Issue to port to V2, with a link to the current prototype code:
https://github.com/openfoodfacts/openfoodfacts-hungergames/issues/47
Β 
== Form recognizer ==
Β 
* Previous work on Form Recognizer by Ramzi: https://github.com/openfoodfacts/openfoodfacts-ai/tree/master/nutrition-table-extraction/notebooks
* Can we do the form recognizer query on the fly from the page ?
* Should we do it server side ?
* We already have an OFF API for extraction nutrition data from text that we can also use.


[[Category:Hackathons]]
[[Category:Hackathons]]
[[Category:Artificial Intelligence]]
[[Category:Artificial Intelligence]]

Latest revision as of 09:26, 22 August 2024

Back to Nutrition facts table data extraction

Introduction

A team of Microsofties were working during summer 2020 on training custom model to extract data from nutrition facts table pictures to pre-populate Open Food Facts' Hunger Game data entry and validation form.

The results were not satisfactory, and the model was not integrated production.

More recent work on nutrition data extraction is done on the on-going NutriSight project.