metadata

language:
  - en
license: apache-2.0
task_categories:
  - object-detection
dataset_info:
  features:
    - name: image_id
      dtype: int64
    - name: image
      dtype: image
    - name: bboxes
      sequence:
        sequence: float64
    - name: category_ids
      sequence: int64
    - name: category_names
      sequence: string
  splits:
    - name: train
      num_bytes: 321253578
      num_examples: 423
    - name: test
      num_bytes: 215302392
      num_examples: 282
  download_size: 514442499
  dataset_size: 536555970
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
tags:
  - Document Layout Analysis

HeadlineHunter

HeadlineHunter is a novel Document Layout Analysis Dataset centred on newspapers.

The newspapers currently in the dataset are from The Daily Monitor (Uganda), but we hope to add more as time progresses.

Class Labels

{0: 'Ad',
 1: 'Table',
 2: 'byline',
 3: 'caption',
 4: 'deck',
 5: 'folio',
 6: 'headline',
 7: 'illustration',
 8: 'jumpline',
 9: 'masthead',
 10: 'photograph',
 11: 'story'}

Citation Information

@ONLINE{Headline Hunter,
    author = {William Gitta Lugoloobi},
    title = {Headline HUnter},
    month = {sep},
    year= {2024}
}