Skip to content

Dataset · RadGenome

PMC-VQA: Visual Question Answering Dataset

RadGenome/PMC-VQA

227k question-answer pairs over 149k images for visual question answering training.

What's inside

PMC-VQA contains 227k QA pairs and 149k images. The first version includes train.csv, test.csv, test_clean.csv, and an images.zip archive; the second version includes train2.csv, test2.csv, and images2.zip with non-composite images. The language is not specified in the card.

What you can do with it

  • Train a VQA model to answer questions about images
  • Fine-tune LoRA for visual question answering
  • Evaluate a multimodal model on question-answer pairs
Downloads this month
24K
Likes
81

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("RadGenome/PMC-VQA")