Image Search Engine 🔍

A powerful and efficient image search engine that uses deep learning to find similar images based on their visual content. This project combines state-of-the-art neural networks with vector database technology to enable fast and accurate image similarity search.

Features

• Multiple Model Support: Compatible with various deep learning architectures:
  • ResNet50
  • EfficientNet
  • MobileNetV3
• Vector Database Integration: Uses LanceDB for efficient similarity search
• Batch Processing: Optimized for processing large image datasets
• Web Interface: User-friendly FastAPI-based web application
• Configurable: Easy-to-modify YAML configuration files
• Multi-format Support: Handles various image formats (JPEG, PNG, WebP)

Installation

1. Clone the repository:

git clone https://github.com/yourusername/image-search-engine.git
cd image-search-engine

2. Create a virtual environment (recommended):

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

3. Install dependencies:

pip install -r requirements.txt

Usage

1. Image Ingestion

To index your image dataset:

python ingest.py

This will:

• Scan your dataset directory for images
• Extract features using the specified model
• Store embeddings in the vector database

2. Query Images

To search for similar images:

python query_db.py --image_path path/to/your/query/image.jpg

3. Web Interface

Start the web application:

python -m uvicorn app.app:app --reload

Access the web interface at http://localhost:8000

🔧 Configuration

Configuration files are stored in the configs/ directory. Available configurations:

• animals_efficientnet.yml
• animals_mobilenetv3.yml
• animals_resnet50.yml
• ocean_resnet50.yml
• ocean_resnet50_v2.yml

Example configuration:

MODEL_NAME: resnet50
MODEL_DIM: 2048
COLLECTION_NAME: images
DATASET_PATH: ./dataset
LANCEDB: ./vectordb

Project Structure

image-search-engine/
├── app/                    # Web application
├── configs/                # Model configurations
├── dataset/               # Image dataset directory
├── engine/                # Core search engine components
├── vectordb/              # Vector database storage
├── ingest.py              # Dataset ingestion script
├── query_db.py            # Image query script
└── requirements.txt       # Project dependencies

How It Works

1. Feature Extraction: Deep learning models convert images into high-dimensional feature vectors
2. Vector Storage: Features are normalized and stored in LanceDB
3. Similarity Search: Query images are processed the same way and compared using cosine similarity
4. Result Ranking: Most similar images are returned based on vector similarity

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.