High-quality synthetic text data generation for Urdu Text Recognition
Released as a supplement of UTRNet: High-Resolution Urdu Text Recognition
-
Create dataset using the following command
python3 run.py --count 1000 --length 10 --max_length 100 --random --name_format 2 --height 128 --thread_count 8 --skew_angle 10 --random_skew --blur 2 --random_blur --salt_and_pepper 0.1 --distorsion 3 --distorsion_orientation 2 --background 3 --random_fit --random_resize --random_crop --random_shearx --random_margins --margins 5,5,5,5 --output_dir 1k_images -
Run
python3 run.py --helpfor help
- Download the UTRSet-Synth dataset
- For more information & other resources, visit Project Webpage
- Main codebase - UTRNet Repo
- Based on the trdg library
- The code (& the generated dataset) is for research purposes only and must not be used for any other purpose without the author's explicit permission.
If you use the code/model/dataset, please cite the following paper:
@article{rahman2023utrnet,
title={UTRNet: High-Resolution Urdu Text Recognition In Printed Documents},
author={Abdur Rahman and Arjun Ghosh and Chetan Arora},
journal={arXiv preprint arXiv:2306.15782},
year={2023},
eprint={2306.15782},
archivePrefix={arXiv},
primaryClass={cs.CV},
doi = {https://doi.org/10.48550/arXiv.2306.15782},
url = {https://arxiv.org/abs/2306.15782}
}