README.md

metadata

dataset_info:
  features:
    - name: image
      dtype: image
    - name: image_url
      dtype: string
    - name: caption
      dtype: string
    - name: clip_score
      dtype: float64
  splits:
    - name: train
      num_bytes: 1435616055
      num_examples: 500
  download_size: 1435214603
  dataset_size: 1435616055
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Conjunto redizudo de imagens com captions geradas pelo modelo VerboVision

Imagens retiradas do conjunto VerboVision-100K, que é baseado no LAION