README.md

metadata

dataset_info:
  features:
    - name: image
      dtype: image
    - name: instruction
      dtype: string
    - name: answer
      dtype: string
    - name: unnormalized_box
      sequence: float32
    - name: elem_text
      dtype: string
    - name: elem_role
      dtype: string
    - name: func
      dtype: string
    - name: image_size
      dtype: string
    - name: device
      dtype: string
  splits:
    - name: train
      num_bytes: 20372037
      num_examples: 100
  download_size: 20069711
  dataset_size: 20372037
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

This miniset of the AutoGUI-v1 dataset demonstrates representative samples of functionality grounding/captioning tasks.