SuitAI Search Engine

SuitAI-Search-Engine

SuitAI-Search-Engine is an innovative AI-powered search engine designed to revolutionize e-commerce by integrating advanced Natural Language Processing (NLP) and Computer Vision (CV) technologies.

Demo

Check out our project demonstration video:

Project Overview

This project aims to enhance the e-commerce search experience by:

• Improving search accuracy and fairness
• Integrating multiple data types (text and images)
• Enhancing user experience through personalized recommendations
• Enabling innovative product discoverability

Key Features

• Advanced image and text analysis for accurate product matching
• Vector database for efficient similarity search
• Personalized recommendation system
• Image captioning using LLaVA model
• Product image generation using Stable Diffusion
• Multimodal search capabilities (text and image inputs)

Technologies Used

• CLIP (Contrastive Language-Image Pretraining)
• LLaVA (Large Language and Vision Assistant)
• Stable Diffusion
• Vector Database (ChromaDB)
• Streamlit (for user interface)

Project Structure

• App/: Contains the main application code
• CLIP/: CLIP model implementation
• Data Collection/: Scripts for data collection and preprocessing
• Data Format/: Data formatting utilities
• LLaVA/: LLaVA model implementation
• Processing the data/: Data processing scripts
• diffusers/train text to image/: Image generation model training
• recommednation/: Recommendation system implementation
• vectorDB/: Vector database setup and queries

Future Work

• RAG-Based Search
• Implementation of 3D model visualization
• Development of an advanced styling expert system
• Integration of a chatting system for interactive shopping assistance

Contributors

• Mohammed Naseif
• Abdulrahman Hassan
• Mohammed Mufleh
• Nawaf Binsaad
• Abdulaziz Khoja

Acknowledgments

Special thanks to Dr. Soumaya Chaffar for supervision and to the Artificial Intelligence Department at the University of Prince Muqrin for their support.

License

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