Adding an higher level API for experimental ML extraction of nutrition

[teolemon]

Who for

• Smoothie devs

What

• The API is too low level
• Adding an higher level API for experimental ML extraction of nutrition
• From:
  • https://robotoff.openfoodfacts.org/api/v1/predict/nutrient?ocr_url=https://world.openfoodfacts.org/images/products/802/509/314/0251/4.json
• To
  • https://robotoff.openfoodfacts.org/api/v1/predict/nutrients?8025093140251&image=nutrition_fr
• Note: we don't seem to generate JSONs for nutrition_fr
  • https://world.openfoodfacts.org/images/products/541/067/300/4468/nutrition_fr.5.full.jpg

Why

• We could have an initial auto-extraction mode

Part of

• Nutrition (tracker) #372

[raphael0202]

We currently lack a good model for information extraction, after analysis the current approach doesn't work well for nutritional table. So it's better to add this once we have an accurate model.