This project is for a paper called "TEMDnet: A Novel Deep Denoising Network for Transient Electromagnetic Signal with Signal-to-Image Transformation". It is worth pointing out that we don't upload the datasets to this git repository due to the datasets are connected with the secrecy of project. Luckily, we will provide overall datasets in the future. If you have questions,please feel free to contact me. Finally, my Email is [email protected] , which is often uesed.
Recently,data-driven denoising methods have achieved impressive perfromace, where deep Convolutional Neural Networks (CNN) based natural image denoising methods is the most booming field. Thus, an interesting question may be existed, i.e., Can we remove the noise from acquired TEM signal by the state-of-the-art image denoising methods? To this end, we first achieve a novel signal-to-image transformation method in order to keep the structual information of TEM signal as much as possible. We then propose a novel CNN-based denoiser to modeling the noise.
- signal-to-image transformation method
- the proposed CNN-based denoiser
An example dataset, including training data and test data, can be downloaded at here
Please refer to the paper for more details and you can also utilize a data_generator script at here
- python 3+
- tensorflow 1.10