Dataset · 5CD-AI
60k lines of handwritten Vietnamese text for OCR
5CD-AI/VietHTR-Line
Manually verified image and transcription pairs of handwritten Vietnamese lines for OCR tasks.
What's inside
Contains 60,247 image-text pairs cropped line by line from real handwritten documents in Vietnamese. The sample features various writing styles ranging from neat handwriting to faded ink and connected strokes. All text transcriptions were verified by humans while fully preserving Vietnamese diacritics.
What you can do with it
- Training and fine-tuning handwritten text recognition (HTR / OCR) models for the Vietnamese language
- Testing multimodal models on complex real-world documents
- Evaluating diacritic recognition accuracy
- Tasks
- Image → text
- Languages
- Vietnamese
- Size
- 10K–100K rows
- Downloads this month
- 517
- Likes
- 25
- Access
- on request on Hugging Face
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("5CD-AI/VietHTR-Line")