AI-Powered Image Search Engine

A modern image search engine that uses CLIP (Contrastive Language-Image Pre-Training) for semantic search and Qdrant for vector similarity search. This system allows you to search images using either text descriptions or similar images.

Features

• 🖼️ Auto-Indexing: Automatically indexes images from the selected folders
• 🔍 Text Search: Find images using natural language descriptions
• 📸 Image Search: Find similar images by uploading a reference image
• ⚡ Real-time Updates: Monitors the all folders for new images and indexes them automatically
• 🎯 High Accuracy: Uses OpenAI's CLIP model for high-quality image-text matching
• 📊 Similarity Scores: Shows match percentage for each result

Requirements

• Python 3.8+
• CUDA-compatible GPU (optional, but recommended for better performance)

Installation

1. Clone the repository:

git clone https://github.com/itsfuad/SnapSeek
cd SnapSeek

2. Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

Usage

1. Start the backend server:

python main.py

2. Open your browser and navigate to:

http://localhost:8000

3. Add your photo folders and start searching!

Development

Running Tests

1. Install test dependencies:

pip install -r requirements-test.txt

2. Run the tests:

pytest tests/ -v

GitHub Actions

The project includes GitHub Actions workflows for automated testing:

• Tests run on every push to main branch
• Tests run on every pull request to main branch
• Uses Python 3.11 on Ubuntu latest
• Installs dependencies from both requirements.txt and requirements-test.txt

Project Structure

Project
├── main.py              # FastAPI application entry point
├── requirements.txt     # Main dependencies
├── requirements-test.txt # Test dependencies
├── templates/          # Frontend templates
├── static/            # Static assets
├── tests/             # Test files
└── .github/
    └── workflows/     # GitHub Actions workflows

2. Poor Search Results:

   • Try more specific search queries
   • Add more diverse images to the dataset
   • Use image search for more precise matching

3. Performance Issues:

   • Consider using GPU acceleration
   • Reduce the number of indexed images
   • Adjust the similarity threshold

License

This project is licensed under the Mozilla Public License 2.0 - see the LICENSE file for details.

Credits

• CLIP model by OpenAI
• Qdrant vector database
• FastAPI framework