Image Similarity Clustering Engine

This Streamlit application helps you discover visually similar groups within your image collection using MediaPipe's Image Embedder and hierarchical clustering. Upload your images, and the application will automatically group them based on visual similarity.

Key features:

• Support for up to 20 images (10MB each)
• Interactive similarity threshold adjustment
• Expandable group results view
• Powered by MediaPipe's MobileNet V3 model

Built With

• Streamlit
• MediaPipe
• OpenCV
• Python
• SciPy (Hierarchical Clustering)

App Screenshots

Prerequisites

The application requires the MediaPipe MobileNet V3 model. Download it using:

powershell -Command "Invoke-WebRequest -Uri https://storage.googleapis.com/mediapipe-models/image_embedder/mobilenet_v3_small/float32/1/mobilenet_v3_small.tflite -OutFile models/mobilenet_v3_small.tflite"

Local Setup

1. Clone the repo

   git clone https://github.com/pedropcamellon/image-similarity-clustering-engine.git
   cd image-similarity-clustering-engine

2. Install dependencies using uv

   uv sync

3. Run the Streamlit app

   streamlit run src/app.py

Usage

1. Open the application in your web browser
2. Upload 2-20 images using the file uploader
3. Adjust the similarity threshold if needed
4. Click "Start Clustering"
5. View the results in expandable group sections

Technical Details

Core Components

1. MediaPipe Image Embedding Pipeline

   • Uses MobileNet V3 for feature extraction

2. Clustering Engine

   • Hierarchical clustering with customizable threshold
   • Cosine similarity measurements
   • Automatic group detection

Performance Considerations

• Caching strategy using @st.cache_data
• Efficient image processing with OpenCV
• Upload limits for optimal performance
  • Maximum 20 images
  • 10MB per image

License

This application is released under the GNU General Public License.