This repository contains a Jupyter Notebook that implements a seismic data denoising technique using a Vision Transformer (ViT). It leverages machine learning and deep learning frameworks to effectively remove noise from seismic signals, improving data quality for further geophysical analysis.
- Utilizes Vision Transformer (ViT) for seismic data denoising.
- Supports GPU acceleration for faster training and inference.
- Implements machine learning and deep learning techniques using TensorFlow.
- Includes visualization tools for assessing denoising performance.
- Supports data preprocessing, training, evaluation, and model saving/loading.
Ensure you have the following libraries installed before running the notebook:
- Python (>= 3.7)
- TensorFlow
- scikit-learn
- NumPy
- Matplotlib
- Pickle
Install the required dependencies using:
pip install tensorflow scikit-learn numpy matplotlib pickle
This notebook is optimized for GPU usage. Ensure you have the proper CUDA and cuDNN versions installed along with TensorFlow to utilize GPU acceleration.
git clone https://github.com/SimLab120/SeisDenoiser.git
cd ./SeisDenoiser
Ensure your seismic dataset is correctly formatted and loaded Look at the codebase to know appropriate shape of seismic section image to reshape your own seismic section data and Modify padding accordingly.
For reasons, Data folder used for our codebase is not provided. Do look into Data Loader section of notebook to know more about how Data is organized.
Execute all cells to train and evaluate the model.
Visualize the denoised seismic data and compare it with the noisy input.
Use provided functions to save and load trained models.
Contributions are welcome! Feel free to fork this repository, create a new branch, and submit a pull request with your improvements.
For queries, contact [email protected].