Content & Text Based Image Retrieval Search Engine

Introduction | Functionalities | Model Architecture | Results | Deployment | Tools | Installation | Topics

🔗Introduction

An image search engine that covers over 1 million annotated images from Open Images Dataset. Users can conveniently search using text, images, or a combination of both.

🔗Functionnalities

This project employs Elasticsearch to store data and provides two primary functions:

1. Similarity Search: It retrieves the most similar images to a provided reference image.
2. Database search: It conducts searches based on image tags, enabling users to find images that match specific descriptors.

🔗Model Architecture

🔗Results

Below you can find some of the results:

🔗 Deployment

Both the API and frontend of the web application are efficiently containerized using Docker and docker-compose. For more details, refer to the comprehensive report.

🔗Tools

• Frontend Development: Streamlit
• Backend Development: Flask
• Database: Elasticsearch 8.1.0 + Elasticknn plugin 8.1.0 (You can install it from this link: https://github.com/alexklibisz/elastiknn/releases/tag/8.1.0.0)
• Deployment: Docker

🔗Installation

In order to reproduce the project you have to:

1- Clone the repository.
2- Move to the project directory.
3- Get the data index.
    * Unzip data fromthe folder data/photo_metadat.zip
    * Change paths in the file models/create_data.py and run it to create feature vectors 
    * Create a new index: execute the file create_index_flickrphotos.py
    * Execute the file mapping.py
4- Run docker-compose up --build Once these steps are done, the web application should start in the browser.
5- You can also execute the app.py.

🔗Topics:

1- Backend Development, deployment
2- Database Managment