FoodCLS

FoodCLS is to map product category tags from OpenFoodFacts to a set of predefined, user-friendly food categories.

The model uses the categories_tags attribute from the OpenFoodFacts product database to classify products into 12 broader food categories.

This allows for simplified product categorization, easier data aggregation, and more intuitive overviews of food products.

Dataset

The model is based on the OpenFoodFacts Product Database, using the state from November 9, 2026.

Target Classes

The model predicts one of 12 predefined categories. Each category is assigned a numeric class ID.

Class ID	German Category	English Equivalent
0	Obst & Gemüse	Fruits & Vegetables
1	Fleisch, Fisch & Alternativen	Meat, Fish & Alternatives
2	Milchprodukte & Eier	Dairy Products & Eggs
3	Brot & Backwaren	Bread & Bakery Products
4	Nudeln, Reis & Getreide	Pasta, Rice & Grains
5	Fertigprodukte	Ready-Made Meals & Convenience Foods
6	Backzutaten	Baking Ingredients
7	Getränke	Beverages
8	Snacks & Süßigkeiten	Snacks & Sweets
9	Öle, Gewürze & Saucen	Oils, Spices & Sauces
10	Brotaufstriche	Spreads
11	Sonstiges	Other

Usage

1. Grab files

First, download the food_classifier.py and requirements.txt.
If you want to use only local files (and not rely on internet connection in ordedr to load weights), ensure to download all files and explicitly setting the local_file_only parameter when calling the from_pretrained()method!

2. Install prerequisites

Next, install the libraries required for running the model:

pip install -r requirements.txt

3. Enjoy usage

For example, like this:

from food_classifier import FoodClassifier


classifier = FoodClassifier.from_pretrained("emiloetting/FoodCLS")

category = classifier.predict(
  {"categories_tags": ["en:milks", "en:dairies"]},
  string_classes=True)
print(category)

When setting string_classes, the ouput will not be the integer class encoding but the respective string stored in the model's attribute self.class_mapping (a dictionary containing the integer-string mapping)

About

This model was made for use in a student project, aiming at providing an intuitive overview over food products.