---
dataset_info:
  features:
  - name: name
    dtype: string
  - name: country_label
    dtype: string
  - name: category_labels
    struct:
    - name: plant-based-foods
      dtype: string
    - name: cereals-and-potatoes
      dtype: string
    - name: beverages
      dtype: string
    - name: dairies
      dtype: string
    - name: meats-and-their-products
      dtype: string
    - name: sweet-snacks
      dtype: string
    - name: snacks
      dtype: string
    - name: plant-based-foods-and-beverages
      dtype: string
  splits:
  - name: train
    num_bytes: 8953027
    num_examples: 80273
  - name: validation
    num_bytes: 1119856
    num_examples: 10034
  - name: test
    num_bytes: 1119911
    num_examples: 10035
  download_size: 2655136
  dataset_size: 11192794
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
---
# OpenFood Classification
A subset of [openfoodfacts/product-database](https://huggingface.co/datasets/openfoodfacts/product-database) selecting only a balanced set of countries and food categories.

## Labels
There are 2 main labels:
- Country *single label*: The corresponding country of the food/dish among 8 possible values: `italy`, `spain`, `germany`, `france`, `united-states`, `belgium`, `united-kingdom` and `switzerland`.
- Category *multi-label*: The category it belongs to among 8 possible values: `snacks`, `beverages`, `cereals-and-potatoes`, `plant-based-foods`, `dairies`, `plant-based-foods-and-beverages`, `meats-and-their-products` and `sweet-snacks`.

## Dataset
There are 8 countries and 8 different categories.
Due to the nature of each label, the dataset is split as follows:
- `name`: The name of the food/dish, extracted from the `product_name` of the openfoodfacts/product-database dataset.
- `country_label`: The country ID, extracted from `countries_tags` of the openfoodfacts/product-database dataset.
- `category_labels`: The categories it belongs to, extracted from `categories_tags` of the openfoodfacts/product-database dataset.

### Distribution
![image/png](https://cdn-uploads.huggingface.co/production/uploads/64161701107962562e9b1006/UjF0RWnrYMTkN3SQZCjMl.png)
Note that the categories overlap each other, since a sample can have multiple categories.

### Splits
The dataset was split into 3 sets:
- `train`: 80%
- `validation`: 10%
- `test`: 10%