openpecha/

OCR-Lhasakanjur

A line-to-text dataset for Tibetan OCR.

	Dataset Details




	
	Dataset Structure

Features:

id: Image file identifier image: Image of a line of text label: Text transcription

Splits:

Train: 130,109 samples Eval: 14,877 samples
Test: 16,326 samples

	Uses




	
	Direct Use

Training and evaluation of Tibetan OCR models Multi-script OCR development Large-scale OCR model pretraining

Not… See the full description on the dataset page: https://huggingface.co/datasets/openpecha/OCR-Lhasakanjur.