Datasets:

manifold-machines
/

lnqa-mk

dataset_info:
  features:
    - name: image
      dtype: image
    - name: translated_qa
      list:
        - name: question
          dtype: string
        - name: answer
          dtype: string
  splits:
    - name: train
      num_bytes: 116089575067.618
      num_examples: 250337
  download_size: 113603455198
  dataset_size: 116089575067.618
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: cc
task_categories:
  - visual-question-answering
language:
  - mk
pretty_name: lnqa-mk

This dataset is based on the lnqa dataset by @vikhyatk. It contains 250,000 images with several question-answer pairs per image.

lnqa-mk was machine translated by Manifold Machines using our Synglot library, with Meta's nllb-200-3.3B model as the backend.