Image Search Engine with Weaviate

This project implements a visual similarity search engine using Weaviate vector database and neural network-based image embeddings. It allows you to upload images and find visually similar ones using AI.

Prerequisites

• Docker and Docker Compose
• Node.js v16 or higher
• npm

Setup

1. Clone the repository:

git clone https://github.com/TennysonGeorge/search-image-service.git
cd search-image-service

2. Install dependencies:

npm install

3. Start the Weaviate services:

docker-compose up -d

This will start:

• Weaviate vector database
• ResNet50 image inference API

4. Start the application:

npm start

5. Open your browser and navigate to:

http://localhost:3000

Features

Web Interface

• Drag and drop interface for image uploads
• Visual results display
• Similarity scoring

REST API

Endpoint	Method	Description
/api/status	GET	Check API status
/api/images	GET	List all indexed images
/api/upload	POST	Upload a single image
/api/upload/bulk	POST	Batch upload from img directory
/api/search	POST	Find similar images

API Usage Examples

Upload an Image

curl -X POST -F "image=@/path/to/your/image.jpg" http://localhost:3000/api/upload

Search for Similar Images

curl -X POST -F "image=@/path/to/your/query.jpg" http://localhost:3000/api/search

Batch Upload from Directory

curl -X POST http://localhost:3000/api/upload/bulk

Project Structure

• server.js: Express API server
• utils.js: Shared utility functions
• upload.js: Image upload functionality
• search.js: Image search functionality
• public/: Web interface files
• docker-compose.yml: Defines required services
• img/: Directory for source images to be indexed
• uploads/: Temporary directory for uploaded files
• results/: Directory for search result images

How It Works

1. Image Vectorization: Uses ResNet50 neural network to convert images into high-dimensional vectors
2. Vector Storage: Weaviate stores these vectors in an efficient searchable format
3. Similarity Search: When searching, new images are converted to vectors and compared to find the most similar matches

Troubleshooting

• If you get connection errors, ensure Docker containers are running:

docker-compose ps

• To reset the database, stop and remove containers:

docker-compose down -v

Technologies Used

• Weaviate: Vector database
• Express.js: Web framework
• ResNet50: Neural network for image feature extraction
• Node.js: Runtime environment
• Docker: Containerization

License

ISC