A Kraken-specific binary training dataset of 897,883 rendered line images
for Burmese/Pali text recognition. Stored in Kraken's native PyArrow IPC format
for direct use with ketos train -f binary — no extraction or conversion needed.
This is not a general-purpose image dataset. The file is a Kraken binary
arrow file (PyArrow IPC) with images stored inline as PNG bytes and Kraken-specific
metadata (alphabet, split counts, image mode). It… See the full description on the dataset page:
https://huggingface.co/datasets/pndaza/burmese-kraken-1m.
