laicsiifes/coco-captions-pt-br

🎉 COCO Captions Dataset Translation for Portuguese Image Captioning

💾 Dataset Summary

COCO Captions Portuguese Translation, a multimodal dataset for Portuguese image captioning with 123,287 images, each accompanied by five 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/coco-captions-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=RGB size=640x480>,
  'caption': [
    'Um restaurante possui mesas e cadeiras modernas de madeira.',
    'Uma longa mesa de restaurante com cadeiras de vime com encosto arredondado.',
    'uma longa mesa com uma planta em cima cercada por cadeiras de madeira',
    'Uma longa mesa com um arranjo de flores no meio para reuniões',
    'Uma mesa é adornada com cadeiras de madeira com detalhes em azul.'
  ],
  'url': 'http://images.cocodataset.org/train2014/COCO_train2014_000000057870.jpg',
  'filepath': 'train2014',
  'filename': 'COCO_train2014_000000057870.jpg',
  'sentids': [787980, 789366, 789888, 791316, 794853],
  'imgid': 40504,
  'split': 'train',
  'cocoid': 57870
}

🗃️ Data Fields

The data instances have the following fields:

• image: a PIL.Image.Image object containing image.
• caption: a list of str containing the 5 captions related to image.
• url: a str containing the url to original image.
• filepath: a str containing the path to image file.
• filename: a str containing name of image file.
• sentids: a list of int containing the ordered identification numbers related to each caption.
• imgid: a int containing image identification number.
• split: a str containing data split. It stores texts: train, val, restval or test.
• cocoid: an int containing example identifier in COCO dataset.

✂️ Data Splits

The dataset is partitioned using the Karpathy splitting appoach for Image Captioning (Karpathy and Fei-Fei, 2015). For training, the train and restval splits are put together as an unique training split with 113,287 examples.

Split	Samples	Average Caption Length (Words)
Train	82,783	10.3 ± 2.7
RestVal	30,504	10.3 ± 2.7
Validation	5,000	10.3 ± 2.7
Test	5,000	10.3 ± 2.7
Total	123,287	10.3 ± 2.7

📋 BibTeX entry and citation info

@misc{bromonschenkel2024cocopt,
  title        = {COCO Captions 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/coco-captions-pt-br}},
  publisher    = {Hugging Face},
  year         = {2024}
}