MiniDeep: A Standalone AI-Edge Platform with a Deep Learning-Based MINI-PC and AI-QSR System

Area of Science:

• Artificial Intelligence
• Cloud Computing
• Edge Computing

Background:

• Developing comprehensive deep learning environments for the entire lifecycle remains a challenge.
• Integrating cloud and edge computing for AI applications requires robust platforms.

Purpose of the Study:

• To introduce MiniDeep, a novel AI-Edge platform providing a complete deep learning development environment.
• To implement and evaluate an AI-QSR KIOSK recommendation system using the MiniDeep platform.

Main Methods:

• MiniDeep utilizes a cloud-edge architecture with Amazon Web Services (AWS) and OpenVino.
• A mini deep lifecycle (MDLC) system with microservices manages training, packaging, inference, and data provision.
• An LSTM-based recommendation system was developed and trained using MiniDeep for AI-QSR KIOSK applications.

Main Results:

• The MiniDeep platform successfully facilitated the end-to-end deep learning lifecycle for the AI-QSR application.
• The LSTM-based recommendation system demonstrated superior performance compared to a rule-based system.
• Key performance metrics including purchase hit accuracy, precision, recall, and F1 score showed significant improvements.

Conclusions:

• MiniDeep offers a comprehensive and integrated solution for deep learning development and deployment.
• The AI-QSR KIOSK recommendation system effectively leverages deep learning for enhanced user experience and sales.
• The proposed platform and recommendation system show significant potential for real-world AI applications.