Skip to content

Dataset · ahmed-masry

Chart and graph question-answering dataset

ahmed-masry/ChartQA

Tens of thousands of chart image and text question pairs for training VLM models.

What's inside

A multimodal dataset of 10K to 100K examples combining images and text (judging by the name, for ChartQA tasks). The card states that it cannot be loaded directly via the load_dataset function: the archive with data and images must be downloaded manually from the Files and versions tab. A separate version without images is also available, which can be loaded standardly via the datasets library.

What you can do with it

  • Training and fine-tuning multimodal models (VLM) for chart and diagram comprehension and analysis
  • Benchmarking information extraction systems from visual reports and presentations
Size
10K–100K rows
Downloads this month
9K
Likes
33

How to load it

from datasets import load_dataset

ds = load_dataset("ahmed-masry/ChartQA")