README.md

---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: imagewidth
    dtype: int32
  - name: text
    dtype: string
  - name: mask
    dtype: image
  - name: image_id
    dtype: int64
  - name: ann_id
    dtype: int64
  - name: img_height
    dtype: int32
  - name: img_width
    dtype: int32
  - name: bbox
    sequence: float32
    length: 4
  - name: prompt
    dtype: string
  - name: images
    sequence: image
  - name: answer_json
    dtype: string
  splits:
  - name: train
    num_examples: 120000
configs:
- config_name: default
  data_files:
  - split: train
    path: train-*
license: mit
task_categories:
- question-answering
language:
- en
tags:
- synthetic
size_categories:
- 10K<n<100K
---