PipeForge: DBSCAN Image Clustering Pipeline

PipeForge is a modular C++ pipeline framework for processing images using a DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm. It supports image input via local files or HTTP(S) URLs and uses OpenCV for image manipulation and contour detection.

✨ Features

• 🌐 Download images via URL using libcurl
• 🧹 Preprocess images to extract contours and center points
• 📊 Apply DBSCAN clustering to identify groups of points
• 🎯 Visualize results with cluster-colored overlay and contour image
• 🔌 Modular pipeline design allows flexible processing steps

---

📂 Project Structure

. ├── include/ # Header files ├── src/ # Source files ├── CMakeLists.txt # CMake build script ├── Dockerfile # Optional container setup └── README.md # This file

---

🔧 Dependencies

• C++17
• OpenCV
• Boost.ProgramOptions
• libcurl
• nanoflann (fetched automatically)

---

⚙️ Build Instructions

🔨 Build (with CMake)

git clone https://github.com/your_username/PipeForge.git
cd PipeForge
mkdir build && cd build
cmake ..
cmake --build .


🐳 Optional: Build with Docker

docker build -t pipeforge .
docker run --rm -v $(pwd):/data pipeforge ./ClusteringRows --path /data/your_image.jpg

🚀 Usage
./ClusteringRows --path /path/to/image.jpg
./ClusteringRows --path https://example.com/image.jpg

Output

centerImage.jpg: Clusters drawn as colored circles
contours.jpg: Original image with contours

📷 Example

📐 DBSCAN Parameters
You can configure DBSCAN via:

const int MINIMUM_POINTS = 10;  // Minimum points to form a cluster
const float EPSILON = 150.0f;   // Radius for neighborhood search


Modify these in dbscanpipe.cpp to tune clustering.

🧱 Pipeline Architecture

Each processing step implements the IPipeline interface and links to the next via .next():
downloader->next(preprocess.get())
          ->next(dbscan.get())
          ->next(output.get());

🧑‍💻 Author
İlke Doğan

GitHub: doganilke

📝 License
This project is licensed under the MIT License. See LICENSE for details.