A Real-Time Reconfigurable AI Processor Based on FPGA

As an AI application requires a lot of resources, optimization of hardware and software to target application is essential. Unlikely the software updates, it is hard to reconfigure the hardware architecture to the target application due to the static characteristics. In this paper, a real-time reconfigurable AI processor based on FPGA is proposed. The AI processor includes a reconfiguration block to update the FPGA and enables hardware reconfiguration with reasonable logic resources. The proposed reconfigurable AI processor was successfully implemented, and the feasibility was demonstrated by experimenting with the accuracy in various applications.