A RISC-V based hardware accelerator designed for Yolo object detection system

Abstract

You only look once (YOLO) is a state-of-the-art, real-time object detection system. A hardware accelerator of YOLO is presented using the open source RISC-V core ROCKET as its controller. Extended customized instructions based on RISC-V were proposed for this accelerator. The hardware design was verified using Xilinx Virtex-7 FPGA VC709 and the results show that the accelerator costs about 400ms to finish the Yolo algorithm and is supposed to have higher speed with more computation modules.