FPGA-Based Hardware Acceleration Using PYNQ-Z2

Abstract

A study on the FPGA development board PYNQ-Z2 for hardware acceleration is presented in this research paper. The experiment accelerates the tasks of optical character recognition (OCR) and image recognition using the FPGA on PYNQ-Z2. The output results on hardware acceleration (Processing system (PS) and Programmable Logic (PL)) are compared with the output results obtained while executing the same tasks on the Arm processor (Processing System (PS)) alone. In this experiment, a Long short-term memory (LSTM) neural network is used to implement OCR, and a Binarized neural network (BNN) is used to implement image recognition. LSTM and BNN here are quantized to reduce memory usage while implementing them on PYNQ-Z2.