Skip to content

Dataset · lmms-lab-encoder

TextVQA Image Text Reading Benchmark

lmms-lab-encoder/textvqa

Formatted version of the TextVQA benchmark for evaluating multimodal models in lmms-eval.

What's inside

Contains pairs of images and QA pairs (from 10k to 100k records) requiring text recognition on images. This is a specially prepared version of the classic TextVQA dataset, adapted for the lmms-eval testing pipeline.

What you can do with it

  • Evaluating the ability of multimodal models (LMM) to read and understand text on images
  • Comparing models in visual question answering tasks with OCR elements
Size
10K–100K rows
Downloads this month
48.8K
Likes
30

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("lmms-lab-encoder/textvqa")