Guided Pyraug

This guide provides instructions for setting up and using Guided Pyraug, a tool for generating latent maps and performing guided sample generation leveraging Pyraug.

---

Installation

1. Create a Conda Environment:

   conda create -n guidedPyraug python==3.10.16

2. Activate the Environment:

  conda activate guidedPyraug

3. Install Required Packages:

    pip install pyraug torch torchvision pandas matplotlib scikit-learn

Creating Latent Maps

Requirements: • Oasis Dataset: NIfTI .nii.gz files. • Oasis Labels: CSV file containing info about mri scans and patients age located in the dataset root directory. • Pretrained Pyraug Weights: A folder containing the trained Pyraug model.

Command:

Run the following script to create a latent map:

  python3 utils/create_latent_map.py --path_to_model_folder <path_to_pyraug_model> --volumes_folders <path_to_oasis_data_samples>

• Replace <path_to_pyraug_model> with the directory path of the Pyraug pretrained model.
• Replace <path_to_oasis_data_samples> with the directory path containing the Oasis dataset.

Output: • The script will generate a .pkl file containing the latent map.

Visualizing the Latent Space:

Use the following script to visualize the latent space:

    python3 utils/show_latent_map.py

Example visualization

Guided Sample Generation

Use the following command to launch the sample generation:

    python3 guided_generation.py --path_to_model_folder <path_to_pyraug_model> --desired_age <float_or_int>

Notes:
	•	Replace <path_to_pyraug_model> with the directory path of the pretrained Pyraug model.
	•	Replace <float_or_int> with the desired age for sample generation.
	•	Additional arguments can be passed; refer to the script for more options.