Customizable U-Net model to be used with any .npy image, with a lot of options to change the internal workings, such as pooling layers used, data augmentation and more settings.
For this project you should have Python 3.9+ installed and preferably CUDA to significantly increase the speed of training.
Clone the respotitory to your desired location.
In order to set up manually the virtual environment you need to run the following commands.
-
Run
python -m venv unet
to create the virtual environment. -
Nagivate into
unet/Scipts
and run theactivate
script. -
Go to the Pytorch website and find the suitable Pytorch version and install it by running the command provided.
-
Run
pip install openpyxl opencv-python
to have all required packages.
Nagivate into the scripts folder in the virtual environment and use activate
if not already active.
Check the config.py file to change any settings required such as data location and result location.
Use command python model.py
to run the model.