Description

Introduction

In this experiment deep learning-based Object recognition using Python is done in PYNQ Z2 Board, which has ZYNQ as a core by Xilinx. Python program can be developed and deployed using Jupyter Notebook in this Board by having SD Card interface with PYNQ-Z2 booted image.

Abstract

In this Object recognition technique, Tiny Yolo Model is used, which is the pre-trained model trained to recognize multiple objects. The below figure is the Architecture of Tiny Yolo.

Existing System

In the existing system, the development of Image/Video-based application in FPGA is complex. But, by using python programming that application can be easily developed and deployed in real-time with the help of Jupyter Notebook in this FPGA board.

Proposed System

In this proposed system, Xilinx’s ZYNQ on PYNQ boards helps to run a Python program in FPGA, here deep learning application for recognizing object is deployed by using Tiny Yolo a pre-trained model.

Connection description

There are two methods for setting up PYNQ. Here we used the 2^nd method by connecting PYNQ with the router through Ethernet cable. Booted SD Card is inserted to the PYNQ board to make it support Python programming through Jupyter Notebook. USB Camera is connected to the USB port of the PYNQ board.

Project description

In this project, you will learn to program FPGA with python to classify the object present in the image frame acquired by the camera connected to the PYNQ. YOLO – You Only Look Once is Fully convolutional Network and output are obtained by Feature map with 1×1 Kernel. Here HDMI Monitor is connected with PYNQ board, displays result frame for every two seconds in the Monitor.

Hardware required

• PYNQ Z2
• Laptop/PC
• USB Camera

Software required

• PYNQ Z2 boot image
• SD Card Formatter
• Etcher/Win32 disk imager

Result

This project classifies the object present in an image using YOLO, that the resulting image can be seen in the Jupyter Notebook. Every program files and result and network model is stored in SD Card, it can be retrieved from the jupyter notebook.