CLIP Image Search Backend

A backend for a search engine that uses the CLIP model and Qdrant vector database to match text queries with relevant images. Ideal for product discovery and cross-modal search applications.

---

Features

• Converts text queries into image-space embeddings.
• Retrieves images that best match the textual description.
• Built using CLIP and Qdrant for high performance.

---

Installation

1. Clone the repository:

   git clone https://github.com/tandalalam/CLIP-image-search-backend.git  
   cd CLIP-image-search-backend/src  

2. Install dependencies:

   pip install -r requirements.txt  

3. Set up a Qdrant instance and configure the .env file:

   QDRANT_HOST=<your_qdrant_host>:<port>  

4. Run the server:

   python main.py  

You can also run the project using Docker by simply docker build -t clip-search and then docker run -p 8080:8080 clip-search.

---

Usage

Send a GET request with a text query to retrieve matching images:

curl -X GET -H "Content-Type: application/json" -d '{"query": "red shoes"}' http://localhost:8080/search