A Streamlit application that uses Qdrant vector database and CLIP model for text-to-image matching.
- Upload multiple images through a simple web interface
- Search images using natural language queries
- Real-time similarity matching using CLIP embeddings
- Persistent storage using local Qdrant instance
- Visual results display with similarity scores
- Python 3.8+
- Qdrant running locally on port 6333
- Docker (optional, for running Qdrant)
-
Clone this repository
-
Install dependencies:
pip install -r requirements.txt- Start Qdrant (if not already running):
# Using Docker
docker run -p 6333:6333 qdrant/qdrant- Start the Streamlit app:
streamlit run app.py-
Open your browser and navigate to the URL shown in the terminal (typically http://localhost:8501)
-
Upload images using the sidebar uploader
-
Enter text queries in the search box to find similar images
.
├── README.md
├── requirements.txt
└── app.py
-
Image Processing:
- Images are uploaded through the Streamlit interface
- CLIP model encodes images into vector embeddings
- Vectors are stored in Qdrant along with the image data
-
Text Search:
- User enters a text query
- CLIP model converts text to vector embedding
- Qdrant performs similarity search
- Most similar images are displayed with scores
- Uses CLIP ViT-B-32 model for encoding both images and text
- Vector dimension: 512
- Distance metric: Cosine similarity
- Images are stored as base64 encoded strings in Qdrant payloads
The app connects to Qdrant using these default settings:
- Host: localhost
- Port: 6333
To modify these settings, update the init_qdrant() function in app.py.
- Currently only supports JPG, JPEG, and PNG image formats
- In-memory image processing may be limited by available RAM
- Search results are limited to top 5 matches
Feel free to open issues or submit pull requests for any improvements.
[Your chosen license]