README.md

metadata

license: apache-2.0
task_categories:
  - image-to-text
  - question-answering
  - zero-shot-classification
language:
  - en
multilinguality:
  - monolingual
task_ids:
  - text-scoring
pretty_name: HL (High-Level Dataset)
size_categories:
  - 10K<n<100K
annotations_creators:
  - crowdsurced
annotations_origin:
  - crowdsurced
dataset_info:
  features:
    - name: file_name
      dtype: string
      captions:
        - name: scene
          sequence:
            dtype: string
        - name: action
          sequence:
            dtype: string
        - name: rationale
          sequence:
            dtype: string
        - name: object
          sequence:
            dtype: string
  splits:
    - name: train
      num_examples: 13498
    - name: test
      num_examples: 1499