Dataset · vankey
Multilingual benchmark for document forgery detection
vankey/RealText-V2
Over 20K document images across six languages for forgery detection and fake text analysis.
What's inside
A large benchmark containing over 20,000 document images in English, Chinese, Arabic, Thai, Malay, and Indonesian. The dataset is designed for detecting text forgeries in images and supports segmentation, classification, and visual question answering (VQA) tasks.
What you can do with it
- Training systems to verify the authenticity of document scans and photos
- Testing models on the segmentation of edited text regions
- Tasks
- Image segmentation, Text classification, Visual question answering
- Languages
- English, Chinese, Arabic, Thai, Malay, Indonesian
- Size
- 10K–100K rows
- Downloads this month
- 44.7K
- Likes
- 8
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("vankey/RealText-V2")