An Optimized SSD Framework for Tube Color Detection on Ultra-low-Power Edge Devices

With the growing demand for intelligent and responsive consumer electronics, deploying AI directly on edge devices has become a key enabler for next-generation smart systems. This paper presents a lightweight color object detection system based on a lightweight Single Shot MultiBox Detector (SSD) architecture, optimized for low-power edge platforms. The proposed model is implemented and evaluated on devices such as the MAX78000, STM32H7, STM32L4, and NXP i.MX 8M Plus, achieving real-time inference under severe resource constraints. Unlike cloud-based solutions, the system performs entirely on-device, ensuring low latency, high energy efficiency, and enhanced data privacy. The results demonstrate the feasibility of integrating deep learning into portable consumer applications such as smart toys and interactive entertainment systems, highlighting the potential of Edge AI in advancing intelligent and sustainable consumer electronics.