Dataset · sarkau
Infographic and Diagram Text Translation Benchmark
sarkau/ImgMT-Dataset-For-Evaluating-Text-Image-Translation
A multilingual collection of Wikimedia vector diagrams and maps with text coordinate annotations for evaluating image translation.
What's inside
The dataset contains 10,000 to 100,000 examples consisting of infographics, maps, scientific diagrams, and educational illustrations collected from Wikimedia Commons. A key feature is that visually identical vector images differ only in their embedded text labels. Bounding boxes and localized text are provided at identical coordinates for each object.
What you can do with it
- Evaluating the quality of text translation directly on images
- Training OCR models and text localization on complex diagrams and maps
- Fine-tuning image generation and editing models while preserving layout
- Tasks
- Image text → text, Image text → image, Translation
- Size
- 10K–100K rows
- Downloads this month
- 39.1K
- Likes
- 0
How to load it
from datasets import load_dataset
ds = load_dataset("sarkau/ImgMT-Dataset-For-Evaluating-Text-Image-Translation")