---
license: mit
dataset_info:
  features:
  - name: newspaper_name
    dtype: string
  - name: newspaper_arkindex_id
    dtype: string
  - name: page_arkindex_id
    dtype: string
  - name: page_index
    dtype: int64
  - name: page_image
    dtype: image
  - name: zone_arkindex_ids
    sequence: string
  - name: zone_polygons
    sequence:
      sequence:
        sequence: float64
  - name: zone_texts
    sequence: string
  - name: zone_classes
    sequence:
      class_label:
        names:
          '0': HEADER-TITLE
          '1': HEADER-TEXT
          '2': ARTICLE-ILLUSTRATION
          '3': ADVERTISEMENT
          '4': ANNOUNCEMENT
          '5': ARTICLE-TITLE
          '6': ARTICLE-TEXT
          '7': ARTICLE-SUBTITLE
          '8': ARTICLE-INSIDEHEADING
          '9': CAPTION
          '10': AUTHOR
          '11': ARTICLE-TABLE
          '12': SECTION-TITLE
  - name: zone_orders
    sequence: int64
  - name: article_id
    sequence: int64
  splits:
  - name: train
    num_bytes: 911036067.0
    num_examples: 623
  - name: val
    num_bytes: 75099123.0
    num_examples: 50
  - name: test
    num_bytes: 71901518.0
    num_examples: 48
  download_size: 1038604332
  dataset_size: 1058036708.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: val
    path: data/val-*
  - split: test
    path: data/test-*
---

# Newspaper segmentation dataset: Finlam

## Dataset Description

- **Homepage:** [Finlam](https://finlam.projets.litislab.fr/)
- **Point of Contact:** [TEKLIA](https://teklia.com)

## Dataset Summary

The Finlam dataset includes 149 French newspapers from the 19th to 20th centuries. 

Each newspaper contains multiple pages. Page images are resized to a fixed height of 2000 pixels.

Each page contains multiple zones, with different information such as polygon, text, class, and order.

### Split

| set   | images | newspapers |
| ----- | ------:| ----------:|
| train | 623    |  129       |
| val   |  50    |   10       |
| test  |  48    |   10       |

### Languages

Most newspapers in the dataset are French, some English.

## Dataset Structure

### Data Fields

- `newspaper_name`: The name of the newspaper.
- `newspaper_arkindex_id`: The Arkindex element id corresponding to the current newspaper.
- `page_arkindex_id`: The Arkindex element id corresponding to the current page.
- `page_index`: The index of the current page in the current newspaper.
- `page_image`: a PIL.Image object containing the page image. Note that when accessing the image column (using dataset[0]["page_image"]), the image file is automatically decoded. Decoding of a large number of image files might take a significant amount of time. Thus it is important to first query the sample index before the "image" column, i.e. dataset[0]["page_image"] should always be preferred over dataset["page_image"][0].
- `zone_arkindex_ids`: the list of Arkindex element ids corresponding to the zones in the page.
- `zone_polygons`: the list of zone coordinates in the current page.
- `zone_texts`: the list of zone texts in the current page.
- `zone_classes`: the list of zone classes in the current page.
- `zone_orders`: the list of zone indexes in the current page, defining the reading order.
- `article_id`: the list of article indexes defining in which article in the current newspaper the current zone is located.