Skip to content

Dataset · nvidia

NVIDIA synthetic multilingual OCR dataset (with Russian)

nvidia/OCR-Synthetic-Multilingual-v1

Tens of millions of synthetic images for training text recognition in 5 languages, including Russian.

What's inside

Large-scale NVIDIA dataset ranging from 10 to 100 million records, generated using a modified SynthDoG generator for text detection and recognition (OCR) tasks. Supports English, Russian, Chinese, Japanese, and Korean languages; this exact data was used to train the Nemotron OCR v2 model.

What you can do with it

  • Training and fine-tuning text detection and recognition (OCR) models with Russian language support
  • Building document text extraction systems for RAG pipelines
Tasks
Object detection, Image → text
Languages
English, Japanese, Korean, Russian, Chinese
Size
10M–100M rows
Downloads this month
17.5K
Likes
62

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("nvidia/OCR-Synthetic-Multilingual-v1")