metadata

language:
  - pt
size_categories:
  - 10K<n<100K
task_categories:
  - text-to-image
  - image-to-text
  - text-generation
pretty_name: nocaps Portuguese Translation
dataset_info:
  features:
    - name: image
      dtype: image
    - name: image_coco_url
      dtype: string
    - name: image_date_captured
      dtype: string
    - name: image_file_name
      dtype: string
    - name: image_height
      dtype: int32
    - name: image_width
      dtype: int32
    - name: image_id
      dtype: int32
    - name: image_license
      dtype: int8
    - name: image_open_images_id
      dtype: string
    - name: annotations_ids
      sequence: int32
    - name: annotations_captions
      sequence: string
    - name: image_domain
      dtype: string
  splits:
    - name: test
      num_bytes: 3342886710
      num_examples: 10600
    - name: validation
      num_bytes: 1422203749
      num_examples: 4500
  download_size: 4761190122
  dataset_size: 4765090459
configs:
  - config_name: default
    data_files:
      - split: test
        path: data/test-*
      - split: validation
        path: data/validation-*

🎉 nocaps Dataset Translation for Portuguese Image Captioning

💾 Dataset Summary

nocaps Portuguese Translation, a multimodal dataset for Portuguese image captioning benchmark, each image accompanied by ten descriptive captions that have been generated by human annotators for every individual image. The original English captions were rendered into Portuguese through the utilization of the Google Translator API.

🧑‍💻 Hot to Get Started with the Dataset

from datasets import load_dataset

dataset = load_dataset('laicsiifes/nocaps-pt-br')

✍️ Languages

The images descriptions in the dataset are in Portuguese.

🧱 Dataset Structure

📝 Data Instances

An example looks like below:

{
  'image': <PIL.JpegImagePlugin.JpegImageFile image mode=L size=732x1024>,
  'image_coco_url': 'https://s3.amazonaws.com/nocaps/val/0013ea2087020901.jpg',
  'image_date_captured': '2018-11-06 11:04:33',
  'image_file_name': '0013ea2087020901.jpg',
  'image_height': 1024,
  'image_width': 732,
  'image_id': 0,
  'image_license': 0,
  'image_open_images_id': '0013ea2087020901',
  'annotations_ids': [0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
  'annotations_captions': [
    'Um bebê está parado na frente de uma casa.',
    'Uma menina com uma jaqueta branca e sandálias.',
    'Uma criança está parada na frente de uma casa.',
    'Uma criança está vestindo uma camisa branca e parada em uma calçada lateral.',
    'Um garotinho está de fralda e com uma camisa branca.',
    'Uma criança usando fralda e sapatos está na calçada.',
    'Uma criança veste uma camisa de cor clara durante o dia.',
    'Uma criança parada na calçada com uma camisa.',
    'Foto em preto e branco de uma menina sorrindo.',
    'um bebê fofo está sozinho com camisa branca'
  ],
  'image_domain': 'in-domain'
}

🗃️ Data Fields

The data instances have the following fields:

image: a PIL.Image.Image object containing the image.
image_coco_url: a str containing the URL to the original image.
image_date_captured: a str representing the date and time when the image was captured.
image_file_name: a str containing the name of the image file.
image_height: an int representing the height of the image in pixels.
image_width: an int representing the width of the image in pixels.
image_id: an int containing the unique identifier of the image.
image_license: an int representing the license type of the image.
image_open_images_id: a str containing the identifier for the image.
annotations_ids: a list of int containing the unique identifiers of the annotations related to the image.
annotations_captions: a list of str containing the captions describing the image.
image_domain: a str indicating the domain of the image. It can be: in-domain, near-domain or out-of-domain.

📋 BibTeX entry and citation info

@misc{bromonschenkel2024nocapspt,
  title        = {nocaps Dataset Translation for Portuguese Image Captioning},
  author       = {Bromonschenkel, Gabriel and Oliveira, Hil{\'a}rio and Paix{\~a}o, Thiago M.},
  howpublished = {\url{https://huggingface.co/datasets/laicsiifes/nocaps-pt-br}},
  publisher    = {Hugging Face},
  year         = {2024}
}