This project is designed for the segmentation of kerogen images using a semi-supervised learning approach. The code leverages PyTorch and the segmentation_models_pytorch library to implement a deep learning model that is trained on both labeled and unlabeled data.
dataset/kerogens.py: Contains the KerogensDataset class for handling the dataset. util/util.py: Utility functions. util/train_helper.py: Helper functions for training. util/eval_helper.py: Helper functions for evaluation. Main script: The primary script that includes functions for setting the seed, loading data, running training epochs, and managing the training process.
- Clone the Repository
git clone <repository_url> cd <repository_directory>
- Prepare Dataset
Ensure that your dataset is organized as required by the KerogensDataset class. Update the paths in the configuration accordingly.
- Set the Seed:
set_seed(42)
- Get Arguments
Implement the get_args() function to parse necessary arguments such as data directories, save paths, etc.
- Train the model trainer(args, config)
Contact For any issues or questions, please open an issue on the repository or contact me at [email protected]