Skip to content

Dataset · 5CD-AI

Vietnamese OCR and VQA on text-bearing images

5CD-AI/Viet-OCR-VQA

Over 137,000 images with Vietnamese text, descriptions, and question-answer pairs for training VQA and OCR models.

What's inside

The dataset includes over 137,000 images containing Vietnamese text. For each image, a detailed description and 5 questions and answers related to the image text were generated using the Gemini 1.5 Flash model, totaling over 822,679 questions.

What you can do with it

  • Training and fine-tuning (SFT) multimodal models for visual question answering (VQA) in Vietnamese
  • Training scene text recognition (OCR) and document information extraction systems
  • Evaluating visual-language model quality on non-English text data
Tasks
Visual question answering, Image → text
Languages
Vietnamese
Size
100K–1M rows
Downloads this month
120
Likes
45
Access
on request on Hugging Face

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("5CD-AI/Viet-OCR-VQA")