SLKNet: An Attention-based Deep Learning Framework for Downhole Distributed Acoustic Sensing Data Denoising
SLKNet is a supervised-learning-based deep learning framework for denoising the FORGE and SAFOD DAS datasets with different types of noise. It combines the DenseNet and a soft attention mechanism (selective kernel block) to extract different scale features.
If you find this package useful, please do not forget to cite the following paper.
Yang, L., Fomel, S., Wang, S., Chen, X., Chen, Y., and Chen, Y., (2023). SLKNet: An Attention-based Deep Learning Framework for Downhole Distributed Acoustic Sensing Data Denoising, Geophysics, doi: 10.1190/geo2022-0724.1.
BibTeX:
@article{YangDe2023,
title={SLKNet: An Attention-based Deep Learning Framework for Downhole Distributed Acoustic Sensing Data Denoising},
author={Liuqing Yang and Sergey Fomel and Shoudong Wang and Xiaohong Chen and Yunfeng Chen and Yangkang Chen},
journal={Geophysics},
year={2023},
pages={in press},
doi={10.1190/geo2022-0724.1},
}
GNU General Public License, Version 3
(http://www.gnu.org/copyleft/gpl.html)
- Tensforflow-gpu: 2.4.1
- numpy: 1.19.5
- Keras: 2.11.0
- GPU: GeForce RTX 3090 Ti
The FORGE DAS dataset can be downloaded here. The SAFOD DAS dataset can be downloaded here.
You can click here to download the FORGE DAS data, including training and test datasets. Make sure you have the following folder structure in the data directory after you unzip the file:
Data
├──FORGE
├── Train
├── Clean
├── Clean_1.mat
├── Clean_2.mat
.
.
├── Clean_29.mat
└── Clean_30.mat
├── Noisy
├── Noisy_1.mat
├── Noisy_2.mat
.
.
├── Noisy_29.mat
└── Noisy_30.mat
└── Test
├── FORGE_example1.mat
└── FORGE_example2.mat
If you have any suggestions or questions, please contact me:
Liuqing Yang
yangliuqingqin@163.com