/unet

U Net

Primary LanguageJupyter Notebook

UNet

This repository contains an implementation of the U-Net architecture.

Setup

To set up the development environment and configure the sys.path:

  1. Run the init_dev_env.sh script:

    $ source init_dev_env.sh

    This script will create an IPython startup file that adds the project directory to sys.path, allowing you to import modules from this project in Jupyter Notebook.

    Install PyTorch with CUDA support

    The init_dev_env.sh script includes a command to install PyTorch with CUDA 11.7 support:

    $ pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117

    This command installs PyTorch with CUDA 11.7 support, which is optimized for GPU acceleration.

    Note: If your CUDA version is different from 11.7, you should modify this command in the init_dev_env.sh script to match your CUDA version. You can find the appropriate command for your CUDA version on the official PyTorch website: https://pytorch.org/get-started/locally/

  2. Restart Jupyter Notebook for the changes to take effect.

Usage

For usage examples and demonstrations of the UNet implementation, please refer to the unet-example.ipynb notebook in this repository.

Acknowledgements

The UNet implementation used in this project is based on the code shared in the following article:

DACON - U-Net 구현하기

We appreciate the author's contribution to the community by sharing their implementation.

UNet Architecture

Below is a visual representation of the UNet architecture:

UNet Architecture

This diagram illustrates the characteristic U-shaped structure of the UNet, showing the contracting path (left side), bottleneck (middle), and expansive path (right side). The skip connections between the contracting and expansive paths are also clearly visible.