pndaza/burmese-ocr-5m

Burmese OCR 5M

A dataset of 5,104,476 rendered line images for Burmese/Pali text recognition (OCR). General-purpose: usable with any OCR engine (Tesseract, Kraken, Calamari, or custom HTR models).

Splits

split	rows
train	4,082,086
validation	511,435
test	510,955
total	5,104,476

Columns

column	type	description
image	image	PNG line image (struct: bytes + path)
text	string	Transcription (Burmese/Pali)
image_id	string	Unique zero-padded id

Usage

With datasets

from datasets import load_dataset

ds = load_dataset("pndaza/burmese-ocr-5m")
# ds["train"][0]
# → {"image": <PIL.Image>, "text": "...", "image_id": "000000001"}

Extract to files (for Tesseract / Kraken training)

from datasets import load_dataset
from pathlib import Path

ds = load_dataset("pndaza/burmese-ocr-5m", split="train")
out = Path("train"); out.mkdir(exist_ok=True)
for row in ds:
    (out / f"{row['image_id']}.png").write_bytes(row["image"]["bytes"])
    (out / f"{row['image_id']}.gt.txt").write_text(row["text"], encoding="utf-8")

Notes

• The companion dataset pndaza/burmese-kraken-5m ships the same data in Kraken's binary PyArrow IPC format for direct use with ketos train -f binary.
• Line images are rendered, not scanned. The degraded subset is intended to bridge the domain gap to real scans; it does not replace genuine scan data.