Skip to content

Dataset · Voxel51

Combined VQA dataset in unified FiftyOne format

Voxel51/MashUpVQA

12,780 examples for Visual Question Answering (VQA) combined from multiple distinct sources for convenient model evaluation.

What's inside

The dataset contains 12,780 samples for the Visual Question Answering (VQA) task in English. It represents a mix of several existing VQA datasets brought into a unified, consistent FiftyOne library format. Loaded and used directly via FiftyOne tools.

What you can do with it

  • Comprehensive evaluation and comparison of VQA systems and multimodal LLMs on diverse visual questions
  • Debugging and visual error analysis of multimodal models in the FiftyOne interface
  • Validating visual recognition pipelines prior to production
Languages
English
Size
10K–100K rows
Downloads this month
30.8K
Likes
5

How to load it

from datasets import load_dataset

ds = load_dataset("Voxel51/MashUpVQA")