This project hosts the testing code for CharNet, described in our paper:
Convolutional Character Networks
Linjie Xing, Zhi Tian, Weilin Huang, and Matthew R. Scott;
In: Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2019.
pip install torch torchvision
python setup.py build develop
-
Please run
bash download_weights.sh
to download our trained weights. -
For ICDAR 2015, please run the following command line. Please replace
images_dir
with the directory containing ICDAR 2015 testing images. The results will be inresults_dir
.python tools/test_net.py configs/icdar2015_hourglass88.yaml <images_dir> <results_dir>
If you find this work useful for your research, please cite as:
@inproceedings{xing2019charnet,
title={Convolutional Character Networks},
author={Xing, Linjie and Tian, Zhi and Huang, Weilin and Scott, Matthew R},
booktitle={Proceedings of the IEEE International Conference on Computer Vision (ICCV)},
year={2019}
}
For any questions, please feel free to reach:
CharNet is CC-BY-NC 4.0 licensed, as found in the LICENSE file. It is released for academic research / non-commercial use only. If you wish to use for commercial purposes, please contact [email protected].