Segmentation-Digital-Rock-Images

Welcome to the GitHub repository for semantic segmentation of 11 digital rock images from the Digital Rock Portal.

Installing the Requirements and Running the Scripts

1. Setting Up Environment

Before running the scripts, ensure you have the required dependencies installed. You can install them using the following command:

git clone https://github.com/MhmdEsml/Segmentation-Digital-Rock-Images
cd Segmentation-Digital-Rock-Images
pip install -r requirements.txt

This command will install all the necessary Python packages listed in the requirements.txt file.

2. Running Dataset_Download.py

Once you have installed the dependencies you can download and prepare the dataset using the following command:

python Dataset_Download.py

After running this script, you can choose the dataset, the numbers of training, validation, and test data.

These datasets are available:

• Berea
• Bandera Brown
• Bandera Gray
• Bentheimer
• Berea Sister Gray
• Berea Upper Gray
• Buff Berea
• CastleGate
• Kirby
• Leopard
• Parker

Example:

• Please enter the image you want to download (e.g., Berea): Berea
• Please enter the number of train patches: 1000
• Please enter the number of validation patches: 100
• Please enter the number of test patches: 100

3. Running train.py

To train the residual UNET on the selected dataset, you can use the following command:

python train.py

That's it! You have successfully trained a segmentation model on the selected data . Feel free to explore and analyze the output for your digital rock analysis needs.

4. Results

After finishing the training process, you can find the results in ./metrics and ./predictions

Google Colab

You can use the following link to run the code in Google Colab: