A Modified Vision Transformer for Kurdish Cursive RTL Handwritten Text Recognition
Authors
Research Square
April 6, 2026
Abstract
This 2026 Research Square preprint presents a modified Vision Transformer approach for handwritten Sorani Kurdish line recognition and introduces DASNUS. The dataset contains 11,475 lines from 867 writers with writer-separated training, validation and test sets. The authors report 3.47% CER and 17.37% WER on the test set. The work is a preprint and public code or trained-model downloads were not verified.
Keywords
Handwritten Text Recognition
Sorani Kurdish
Vision Transformer
DASNUS
Preprint
Related Datasets
DASNUS
2026
11,475 handwritten lines from 867 writers
JPG, TXT, XLSX
A Sorani Kurdish handwritten-line dataset containing 11,475 lines from 867 writers. The creators define writer-separated training, validation and test splits containing 8,042, 1,700 and 1,733 lines …
Citation
Salih, F. M., & Al-talabani, A. K. (2026). A Modified Vision Transformer for Kurdish Cursive RTL Handwritten Text Recognition. Research Square preprint.