- Full Results Link：https://pan.baidu.com/s/1U83LuXL6UmwYtGDzYaHwkw passwd：WJ45
- pretrained model & test image https://drive.google.com/drive/folders/1xnUTUHphKrrDN3MFJaiwDnN7XYUPq1eM?usp=sharing
Before you run this model,you should install the following packges:
python >= 3.6
tensorflow=2.2.0
argparse
then you should run the model like the following command: CUDA_VISIBLE_DEVICES='0' python main.py --result_path ./result
you can change your CUDA devices id and the path to save result images.
Our paper accepted by eccv workshop.
If you find our paper is useful for you ,
please cite us:
@inproceedings{qian2020bggan,
title={Bggan: Bokeh-glass generative adversarial network for rendering realistic bokeh},
author={Qian, Ming and Qiao, Congyu and Lin, Jiamin and Guo, Zhenyu and Li, Chenghua and Leng, Cong and Cheng, Jian},
booktitle={European Conference on Computer Vision},
}
If you have any issues, please contact mingqian@whu.edu.cn.
Our code can traslate to tflite file in a simple way.
If you wanna reproduce our code, you can use the image pair list in "new_list.txt" for training because we manually clean the train data of EBB! dataset.
If anyone wants to get my unsorted code in an uncommercial way, you can email me & I will send it to you. (I didn't have time to sort it, I can send torch or tf2.0 version)
All rights reserved. Licensed under the CC BY-NC-SA 4.0 (Attribution-NonCommercial-ShareAlike 4.0 International)
The code is released for academic research use only. For commercial use, please contact mingqian@whu.edu.cn.