harperreed / photo-similarity-search

Super simple MLX (apple silicon) CLIP based photo similarity web app

📸 Embed-Photos 🖼️

Welcome to Embed-Photos, a powerful photo similarity search engine built by @harperreed! 🎉 This project leverages the CLIP (Contrastive Language-Image Pre-training) model to find visually similar images based on textual descriptions. 🔍🖼️

🌟 Features

• 🚀 Fast and efficient image search using the CLIP model
• 💻 Works on Apple Silicon (MLX) only
• 💾 Persistent storage of image embeddings using SQLite and Chroma
• 🌐 Web interface for easy interaction and exploration
• 🔒 Secure image serving and handling
• 📊 Logging and monitoring for performance analysis
• 🔧 Configurable settings using environment variables

Screenshot

📂 Repository Structure

embed-photos/
├── README.md
├── generate_embeddings.py
├── requirements.txt
├── start_web.py
└── templates
    ├── README.md
    ├── base.html
    ├── display_image.html
    ├── index.html
    ├── output.txt
    └── query_results.html

• generate_embeddings.py: Script to generate image embeddings using the CLIP model
• requirements.txt: Lists the required Python dependencies
• start_web.py: Flask web application for the photo similarity search
• templates/: Contains HTML templates for the web interface

🚀 Getting Started

1. Clone the repository:

   git clone https://github.com/harperreed/photo-similarity-search.git
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

3. Configure the application by setting the necessary environment variables in a .env file.

4. Generate image embeddings:

   python generate_embeddings.py
   

5. Start the web application:

   python start_web.py
   

6. Open your web browser and navigate to http://localhost:5000 to explore the photo similarity search!

Todo

• Use siglip instead of clip
• add a more robust config
• make mlx optional

🙏 Acknowledgments

The Embed-Photos project builds upon the work of the Apple (mlx!), the CLIP model and leverages various open-source libraries. We extend our gratitude to the authors and contributors of these projects.

Happy searching! 🔍✨