© 2023 NEC Corporation
This repository is an official implementation of our paper "Segmentation-free Direct Iris Localization Networks", WACV2023.
- We publish PyTorch implementation of the iris localization network (ILN) model for evaluation. If you need our pre-trained ILN model, please contact us.
- The output is limited to iris and pupil circles (6 dimensions).
- It can used for research purpose.
- Do Not Use It for Commercial Purpose.
This software is released under the NEC Corporation License. See LICENSE before using the code. If you use our model or codes for your research, please cite following paper.
@InProceedings{Toizumi_2023_WACV,
author = {Toizumi, Takahiro and Takahashi, Koichi and Tsukada, Masato},
title = {Segmentation-Free Direct Iris Localization Networks},
booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
month = {January},
year = {2023},
pages = {991-1000}
}
torch > 1.7.1
numpy > 1.19.2
opencv-python > 4.3.0.36
matplotlib > 3.2.2
- If you need pre-trained model for your research, please contact to us. (Our email address is written in the paper.)
- The NEC license also extends to our pre-trained model. See LICENSE before using our model.
- Prepare pre-trained model by yourself or get it from authors.
- Put your dataset images in "img/<your path>" directory.
- Rewrite following line in test.py to your image directory.
#files = sorted(glob('img/sample/**/*.png', recursive=True))
files = sorted(glob('img/<your path>/**/*.<your extension (png, jpg, ...)>', recursive=True))
- Run "test.py".
- Takahiro Toizumi, NEC Corporation.
- Koichi Takahashi, NEC Corporation.
- Masato Tsukada, Tsukuba University.