ColonINST-v1 Data Card

A large-scale mutlimodal instruction tuning dataset for colonoscopy research. More details refer to our project page: https://github.com/ai4colonoscopy/ColonGPT.

Data description

We introduce a pioneering instruction tuning dataset for multimodal colonoscopy research, aimed at instructing models to execute user-driven tasks interactively. This dataset comprises of 62 categories, 300K+ colonoscopic images, 128K+ medical captions (GPT-4V) generated), and 450K+ human-machine dialogues.

Usage

Our ColonINST-v1 dataset consists of two parts: colonoscopy images and human-machine dialogues (in JSON format). Due to privacy concerns associated with medical images, we are unable to publicly share the original colonoscopy images. Please follow this instruction to prepare all colonoscopy images. If you encounter any difficulties in preparing the data, you can contact us via email (gepengai.ji@gmail.com).

However, the JSON files, which contain the human-machine dialogues, are available here for access. You can follow these steps:

• Make sure you have the datasets library installed. You can install it via pip:

  pip install datasets
  

• You can easily download and load the dataset directly in your Python code:

  from datasets import load_dataset
  
  # Load the ColonINST-v1 dataset
  dataset = load_dataset("ai4colonoscopy/ColonINST-v1")
  
  # Once downloaded, you can explore the dataset like this
  print(dataset)
  

• If you need a specific split (e.g., train, validation, or test), you can access it like this:

  train_dataset = load_dataset("ai4colonoscopy/ColonINST-v1", split="train")
  

License

The content of this project itself is licensed under the Apache license 2.0.