What is UNet?

UNet is a deep convolutional neural network architecture originally designed for biomedical image segmentation. It was introduced in the 2015 MICCAI paper titled:

📄 U-Net: Convolutional Networks for Biomedical Image Segmentation
by Olaf Ronneberger, Philipp Fischer, and Thomas Brox [Paper Link]

UNet's core innovation lies in its U-shaped architecture, consisting of a contracting path (encoder) and an expanding path (decoder) with skip connections that merge feature maps from corresponding encoder and decoder layers. This design allows the network to capture context while achieving precise localization, which is crucial for tasks like cell segmentation, tumor boundary detection, and more.

Why UNet?

UNet’s simplicity and effectiveness made it the go-to architecture for medical image segmentation. Its ability to perform well even on small datasets, its flexibility across 2D and 3D medical imaging tasks, and the ease of implementation have led to hundreds of follow-up works and adaptations over the years. Researchers continue to modify and build upon UNet to improve accuracy, efficiency, and generalization across diverse domains.

🔍 About This Blog

This blog aims to showcase the evolution of the UNet family – a timeline of interesting and influential UNet-based architectures from 2015 to the latest publications. Whether you're a researcher, engineer, or enthusiast, this curated collection serves as a reference point and inspiration for exploring semantic segmentation models built upon UNet.

📚 UNet Family Timeline (updated 15/04/2025)

📅 2015

• U-Net: Convolutional Networks for Biomedical Image Segmentation (MICCAI 2015)
  [Paper] • [PyTorch Implementation] • [Keras]

2016

• V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation [paper] [caffe][pytorch]
• 3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation [paper][pytorch]2017
• H-DenseUNet: Hybrid Densely Connected UNet for Liver and Tumor Segmentation from CT Volumes (IEEE Transactions on Medical Imaging)[paper][keras]
• GP-Unet: Lesion Detection from Weak Labels with a 3D Regression Network (MICCAI) [paper]2018
• UNet++: A Nested U-Net Architecture for Medical Image Segmentation (MICCAI) [paper][my-pytorch][keras]
• MDU-Net: Multi-scale Densely Connected U-Net for biomedical image segmentation [paper]
• DUNet: A deformable network for retinal vessel segmentation [paper]
• RA-UNet: A hybrid deep attention-aware network to extract liver and tumor in CT scans [paper]
• Dense Multi-path U-Net for Ischemic Stroke Lesion Segmentation in Multiple Image Modalities [paper]
• Stacked Dense U-Nets with Dual Transformers for Robust Face Alignment [paper]
• Prostate Segmentation using 2D Bridged U-net [paper]
• nnU-Net: Self-adapting Framework for U-Net-Based Medical Image Segmentation [paper][pytorch]
• SUNet: a deep learning architecture for acute stroke lesion segmentation and
  outcome prediction in multimodal MRI [paper]
• IVD-Net: Intervertebral disc localization and segmentation in MRI with a multi-modal UNet [paper]
• LADDERNET: Multi-Path Networks Based on U-Net for Medical Image Segmentation [paper][pytorch]
• Glioma Segmentation with Cascaded Unet [paper]
• Attention U-Net: Learning Where to Look for the Pancreas [paper]
• Recurrent Residual Convolutional Neural Network based on U-Net (R2U-Net) for Medical Image Segmentation [paper]
• Concurrent Spatial and Channel ‘Squeeze & Excitation’ in Fully Convolutional Networks [paper]
• A Probabilistic U-Net for Segmentation of Ambiguous Images (NIPS) [paper] [tensorflow]
• AnatomyNet: Deep Learning for Fast and Fully Automated Whole-volume Segmentation of Head and Neck Anatomy [paper]
• 3D RoI-aware U-Net for Accurate and Efficient Colorectal Cancer Segmentation [paper][pytorch]
• Detection and Delineation of Acute Cerebral Infarct on DWI Using Weakly Supervised Machine Learning (Y-Net) (MICCAI) [paper](Page 82)
• Fully Dense UNet for 2D Sparse Photoacoustic Tomography Artifact Removal [paper]2019
• MultiResUNet : Rethinking the U-Net Architecture for Multimodal Biomedical Image Segmentation [paper][keras]
• U-NetPlus: A Modified Encoder-Decoder U-Net Architecture for Semantic and Instance Segmentation of Surgical Instrument [paper]
• Probability Map Guided Bi-directional Recurrent UNet for Pancreas Segmentation [paper]
• CE-Net: Context Encoder Network for 2D Medical Image Segmentation [paper][pytorch]
• Graph U-Net [paper]
• A Novel Focal Tversky Loss Function with Improved Attention U-Net for Lesion Segmentation (ISBI) [paper]
• ST-UNet: A Spatio-Temporal U-Network for Graph-structured Time Series Modeling [paper]
• Connection Sensitive Attention U-NET for Accurate Retinal Vessel Segmentation [paper]
• CIA-Net: Robust Nuclei Instance Segmentation with Contour-aware Information Aggregation [paper]
• W-Net: Reinforced U-Net for Density Map Estimation [paper]
• Automated Segmentation of Pulmonary Lobes using Coordination-guided Deep Neural Networks (ISBI oral) [paper]
• U2-Net: A Bayesian U-Net Model with Epistemic Uncertainty Feedback for Photoreceptor Layer Segmentation in Pathological OCT Scans [paper]
• ScleraSegNet: an Improved U-Net Model with Attention for Accurate Sclera Segmentation (ICB Honorable Mention Paper Award) [paper]
• AHCNet: An Application of Attention Mechanism and Hybrid Connection for Liver Tumor Segmentation in CT Volumes [paper]
• A Hierarchical Probabilistic U-Net for Modeling Multi-Scale Ambiguities [paper]
• Recurrent U-Net for Resource-Constrained Segmentation [paper]
• MFP-Unet: A Novel Deep Learning Based Approach for Left Ventricle Segmentation in Echocardiography [paper]
• A Partially Reversible U-Net for Memory-Efficient Volumetric Image Segmentation (MICCAI 2019) [paper][pytorch]
• ResUNet-a: a deep learning framework for semantic segmentation of remotely sensed data [paper]
• A multi-task U-net for segmentation with lazy labels [paper]
• RAUNet: Residual Attention U-Net for Semantic Segmentation of Cataract Surgical Instruments [paper]
• 3D U2-Net: A 3D Universal U-Net for Multi-Domain Medical Image Segmentation (MICCAI 2019) [paper] [pytorch]
• SegNAS3D: Network Architecture Search with Derivative-Free Global Optimization for 3D Image Segmentation (MICCAI 2019) [paper]
• 3D Dilated Multi-Fiber Network for Real-time Brain Tumor Segmentation in MRI [paper][pytorch] (MICCAI 2019)
• The Domain Shift Problem of Medical Image Segmentation and Vendor-Adaptation by Unet-GAN [paper]
• Recurrent U-Net for Resource-Constrained Segmentation [paper] (ICCV 2019)
• Siamese U-Net with Healthy Template for Accurate Segmentation of Intracranial Hemorrhage (MICCAI 2019)

📅 2020

• U²-Net: Going Deeper with Nested U-Structure for Salient Object Detection (Pattern Recognition 2020) [paper] [pytorch]
• UNet 3+: A Full-Scale Connected UNet for Medical Image Segmentation (ICASSP 2020) [paper] [pytorch]
• SA-UNet: Spatial Attention U-Net for Retinal Vessel Segmentation [paper] [pytorch]
• Multi-level colonoscopy malignant tissue detection with adversarial CAC-UNet[paper] [pytorch]

📅 2021

• TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation (arXiv 2021) [paper] [pytorch]
• UNETR: Transformers for 3D Medical Image Segmentation (arXiv 2021) [paper] [pytorch]
• Swin-Unet: UNet-like Pure Transformer for Medical Image Segmentation [paper] [pytorch]

📅 2022

• UNeXt: MLP-based Rapid Medical Image Segmentation Network [paper] [pytorch]
• Focal-UNet: UNet-like Focal Modulation for Medical Image Segmentation [paper] [pytorch]
• Practical Blind Image Denoising via Swin-Conv-UNet and Data Synthesis [paper] [pytorch]

📅 2023

• DA-TransUNet: Dual Attention Transformer U-Net for Improved Segmentation [paper] [pytorch]
• Diffusion Transformer UNet: Medical Image Segmentation with Diffusion Models (MICCAI 2023) [paper]
• Biomedical Image Segmentation Using UNet & Transfer Learning [paper]
• U-Net v2: Rethinking the Skip Connections of U-Net for Medical Image Segmentation [paper] [pytorch]
• Music Source Separation Based on a Lightweight Deep Learning Framework (DTTNET: DUAL-PATH TFC-TDF UNET) [paper] [pytorch]
• Faster Diffusion: Rethinking the Role of the Encoder for Diffusion Model Inference [paper] [pytorch]

📅 2024

• LightM-UNet: Mamba Assists in Lightweight UNet for Medical Image Segmentation [paper] [pytorch]
• SelfReg-UNet: Self-Regularized UNet for Enhanced Feature Discrimination [paper]
• MPU-Net: Modality Preserving U-Net for Multimodal Image Segmentation [paper]
• VM-UNet: Vision Mamba UNet for Medical Image Segmentation [paper] [pytorch]
• Semi-Mamba-UNet: Pixel-Level Contrastive and Pixel-Level Cross-Supervised Visual Mamba-based UNet for Semi-Supervised Medical Image Segmentation [paper] [pytorch]
• SAM2-UNet: Segment Anything 2 Makes Strong Encoder for Natural and Medical Image Segmentation [paper] [pytorch]
• UltraLight VM-UNet: Parallel Vision Mamba Significantly Reduces Parameters for Skin Lesion Segmentation
• LKM-UNet: Large Kernel Vision Mamba UNet for Medical Image Segmentation [paper] [pytorch]
• H-vmunet: High-order Vision Mamba UNet for Medical Image Segmentation [paper] [pytorch]
• ViM-UNet: Vision Mamba for Biomedical Segmentation [paper] [pytorch]

📅 2025

• Multi-scale Conv-Attention U-Net: Adaptive Convolution + Attention for Medical Image Segmentation [paper]
• VMAXL-UNet: Vision Mamba and xLSTM-UNet for Medical Image Segmentation [paper]

If you’re running a Next.js application on an Ubuntu server, you probably want a clean workflow to update your production app every time you push to GitHub. In this guide, we’ll walk through setting up GitHub webhooks, a small server to receive them, and a deployment script that builds and restarts your app with PM2.

✅ Prerequisites

• A public or private Ubuntu server
• Node.js, npm, and PM2 installed
• A running Next.js app managed by PM2
• Your app hosted on GitHub (private or public)
• SSH access to your server

Step 1: Generate SSH Key for GitHub (if needed)

On your Ubuntu server, generate an SSH key so the server can authenticate with GitHub and pull code automatically:

ssh-keygen -t rsa -b 4096 -C "your-github-email@example.com"

Press Enter to accept the default path (~/.ssh/id_rsa).
Now, copy the public key:

cat ~/.ssh/id_rsa.pub

Go to GitHub > Settings > SSH and GPG keys > New SSH key, and paste it in.
Test the connection:

ssh -T git@github.com

You should see a “You’ve successfully authenticated” message.

Step 2: Clone Your Repository

Pick a directory for your app and clone your GitHub repo:

cd /var/www
git clone git@github.com:your-username/your-repo.git
cd your-repo

Install dependencies and build the project:

npm install
npm run build

Start the app with PM2:

pm2 start npm --name "your-app-name" -- start
pm2 save

Step 3: Create the Deployment Script

Create a shell script that will handle updates:

nano ~/deploy.sh

Add this content:

#!/bin/bash

cd /var/www/your-repo

echo "Pulling latest changes..."
git pull origin main

echo "Installing dependencies..."
npm install

echo "Building the app..."
npm run build

echo "Restarting app with PM2..."
pm2 restart your-app-name

echo "Deployment complete!"

Make it executable:

chmod +x ~/deploy.sh

Step 4: Set Up a Webhook Listener

Create a simple Express server that listens for GitHub webhooks:

mkdir ~/webhook && cd ~/webhook
nano server.js

Paste the following code:

const express = require("express");
const { exec } = require("child_process");

const app = express();
app.use(express.json());

app.post("/webhook", (req, res) => {
  const payload = req.body;

  if (payload.ref === "refs/heads/main") {
    console.log("Push to main detected. Starting deployment...");
    exec("bash ~/deploy.sh", (err, stdout, stderr) => {
      if (err) {
        console.error(`Deployment error: ${err.message}`);
        return res.status(500).send("Deployment failed.");
      }
      console.log(stdout);
      console.error(stderr);
      res.status(200).send("Deployment complete.");
    });
  } else {
    res.status(200).send("Push ignored (not main branch).");
  }
});

const PORT = 4000;
app.listen(PORT, () => {
  console.log(`Webhook server listening on port ${PORT}`);
});

Initialize and install dependencies:

npm init -y
npm install express

Start the server using PM2:

pm2 start server.js --name webhook
pm2 save

Step 5: Open the Webhook Port (Optional)

If your firewall is active, allow the webhook port:

sudo ufw allow 4000

Step 6: Set Up GitHub Webhook

In your GitHub repo:

1. Go to Settings > Webhooks > Add webhook
2. Payload URL:

   http://your-server-ip:4000/webhook

3. Content type:

   application/json

4. Leave secret empty (or handle it in code if using)
5. Events: Just the push event
6. Save webhook

Step 7: Test the Deployment Flow

Make a small commit to the main branch:

git add .
git commit -m "Test auto-deploy"
git push origin main

Then monitor the webhook log:

pm2 logs webhook

You should see output indicating that the deployment script was triggered, the app was rebuilt, and restarted via PM2.

What Are the Advantages of Using This Webhook Setup?

1. Hands-Free Deployment
   Every time you push to the main branch, your app automatically updates — no need to SSH into the server, pull manually, or restart services.

2. Fast Feedback Loop
   Code changes go live in seconds, keeping development and production in sync. This is especially useful for small teams or solo developers.

3. No External CI/CD Needed
   You don’t need to rely on tools like GitHub Actions, Jenkins, or CircleCI. Everything runs locally on your own server, which is great for simplicity and control.

4. Customizable Deployment Logic
   You have full control over the deployment process. Want to run tests, update environment variables, or notify Slack? Just add it to your deploy.sh.

5. Lightweight & Resource-Friendly
   Unlike full CI/CD tools that require containers or background jobs, this setup runs with minimal overhead — just a simple Express server and PM2.

6. Works with Private Repos
   Since you're using SSH keys to authenticate with GitHub, this works equally well with private repositories.

7. Secure by Design
   You can lock down access to the webhook port, use GitHub IP whitelisting, and add secret token verification for even more security.

8. Great for VPS and Self-Hosted Environments
   Whether you're using DigitalOcean, Hetzner, Linode, or your own machine, this approach is ideal for traditional Ubuntu-based deployments.

If you're deploying your Next.js 13+ app to GitHub Pages and running into issues where your JavaScript and CSS aren't loading, you're not alone. This issue usually comes down to how GitHub Pages handles static files — especially those in folders that start with an underscore like /_next.

Let me walk you through how to fix it step-by-step. 👇

🧠 The Problem

By default, GitHub Pages uses Jekyll, which ignores folders starting with _, like:

/_next/static/...

This means your app's static assets (JS, CSS) won't be served — resulting in a broken site with missing styles and scripts. 😵

✅ The Solution

1. Add .nojekyll file

This file disables Jekyll processing on GitHub Pages.

You can add this manually in the out folder after next export, or automatically via a script:

touch out/.nojekyll

Or use shx to make it cross-platform compatible (for Windows users):

npm install --save-dev shx

2. Install gh-pages

You’ll need the gh-pages package to deploy the out folder to your GitHub Pages branch.

Install it with:

npm install --save-dev gh-pages

This tool will handle publishing your static site to the correct branch (typically gh-pages or main).

3. Update next.config.js

Configure your app for static export:

/** @type {import('next').NextConfig} */
const nextConfig = {
  output: 'export', // Static export for GitHub Pages
  basePath: '',     // Leave empty for username.github.io
};

module.exports = nextConfig;

If you're deploying to a project page (e.g., username.github.io/my-app), set:

basePath: '/my-app',
assetPrefix: '/my-app',

4. Update your package.json scripts

"scripts": {
  "dev": "next dev",
  "build": "next build",
  "export": "next export",
  "postexport": "shx touch out/.nojekyll",
  "deploy": "next build && npm run postexport && gh-pages -d out --nojekyll --cname https://yourusername.github.io"
}

📝 Replace https://yourusername.github.io with your GitHub URL.

🏁 Deployment Command

npm run deploy

Boom 💥 — your Next.js site should be live and functional!

🔄 GitHub Pages Branch Setup

1. Go to GitHub → Settings → Pages
2. Source: Deploy from branch
3. Branch: gh-pages, folder: / (root)
4. Save ✅

🔗 Bonus Tips

• Use --cname for custom domains
• Use proper basePath if using a project subfolder
• Automate deploys with GitHub Actions

📦 Example Repo

Check out: https://github.com/sky4689524/sky4689524.github.io

🧰 Tools Used

• Next.js 13+
• gh-pages
• shx