https://github.com/muhdhammad/unet-vanilla

Implementation of U-Net architecture in PyTorch
https://github.com/muhdhammad/unet-vanilla

convolutional-neural-networks deep-learning image-segmentation python pytorch unet

Last synced: 2 months ago
JSON representation

Implementation of U-Net architecture in PyTorch

• Host: GitHub
• URL: https://github.com/muhdhammad/unet-vanilla
• Owner: Muhdhammad
• Created: 2025-03-24T16:28:23.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2025-03-24T20:21:51.000Z (over 1 year ago)
• Last Synced: 2025-03-24T20:29:24.264Z (over 1 year ago)
• Topics: convolutional-neural-networks, deep-learning, image-segmentation, python, pytorch, unet
• Language: Python
• Homepage:
• Size: 77.1 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# U-Net: PyTorch Implementation

[![PyTorch](https://img.shields.io/badge/PyTorch-%23EE4C2C.svg?logo=PyTorch&logoColor=white)](https://pytorch.org/)

This repo is a PyTorch implementation of the original U-Net architecture (Ronneberger et al., 2015) for image segmentation. You can view the original paper [here](https://arxiv.org/pdf/1505.04597.pdf).



## Why this implementation?

- Beginner-Friendly - Clear, modular code and easy to understand.

- Flexible Structure - Easy to modify layers, add new features, or experiment with different architectures.

## How It Works

### Contracting Path (left side):

Each downSample block does: 

```

Conv3x3 → ReLU → Conv3x3 → ReLU → MaxPool2x2  

```

### Expanding Path (right side):

Each upSample block does:

```

2x2 Transpose Conv → Concatenate (skip connection) → 3x3 Conv → ReLU

```

Skip connections copy features from left to right to recover spatial details.