Datasets:
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