metadata
dataset_info:
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: conversation
struct:
- name: data
list:
- name: answer
dtype: string
- name: question
dtype: string
splits:
- name: train
num_bytes: 59333908
num_examples: 700
download_size: 59207371
dataset_size: 59333908
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
task_categories:
- question-answering
language:
- en
tags:
- medical
size_categories:
- n<1K
This dataset is made using raw data from MURA (musculoskeletal radiographs), a large dataset of bone X-rays.
For each image in the original raw dataset we used the associated metadata to generate a simulated conversation about the image between a user and a chatbot.