Implementation Research of Tiny ML on RISC-V Platform Using Efinix Platform for Real Time Person Detection

The rapid advancements in artificial intelligence (AI) and computer vision have revolutionized intelligent security systems, addressing challenges related to cost, efficiency, and scalability. This paper investigates the application of Tiny Machine Learning (TinyML) on the RISC-V platform, utilizing the Efinix Ti60 F225 FPGA for real-time person detection. The study aims to harness the energy efficiency, flexibility, and scalability of the RISC-V architecture to develop a cost-effective and sustainable solution for intelligent security applications. The proposed system combines the YOLO model for person detection with TinyML techniques, enabling real-time processing while maintaining an accuracy of 80% for individuals within a 2-meter range. Hardware components, such as the Raspberry Pi Camera Module V2 and Mini DSI Panel, were employed to capture and display real-time imagery. Comprehensive evaluations demonstrated the system's performance advantages, including reduced latency, low power consumption, and minimal bandwidth usage, emphasizing the benefits of deploying AI at the edge. Although challenges such as environmental variability and memory limitations were encountered, the findings highlighted the potential of integrating TinyML with RISC-V platforms for intelligent security applications. The results contributed to the field of edge computing and open-source hardware, presenting a viable approach to energy-efficient and scalable AI solutions. Future research will focus on overcoming hardware constraints, extending the detection range, and incorporating adaptive learning capabilities to enhance system robustness and functionality further.