FaDAS

Team AOHW-154

Team members :

Murat Tokez (m.tokez2019@gtu.edu.tr)

Ilker Acar (i.acar2020@gtu.edu.tr)

Duygu Zeynep Tuncel (d.tuncel2020@gtu.edu.tr)

Mehmet Salih Turhan (m.turhan2020@gtu.edu.tr)

Supervisor:

İhsan Cicek

Introduction

This project is about running and accelerating vehicle detection and mapping ADAS applications in parking lot using Deep Learning Unit (DPU) on KV260 FPGA board.

Documentation

Documentation

Prerequires

• Vitis-AI: High-level libraries and APIs for AI inference with DPU cores.
• Deep Learning Unit (DPUCZDX8G): A configurable computation engine optimized for convolutional neural networks.
• PetaLinux: An embedded Linux SDK targeting FPGA-based SoC designs.

System Architecture

The system includes:

• Kria KV260
• A RPLIDAR A1M8
• A camera
• A rotary encoder and Arduino nano to provide the third axis to the 2D LiDAR, creating a 3D point cloud

Steps Involved:

1. Vivado Project: Create hardware design.
2. Linux Project: Integrate hardware with PetaLinux.
3. Model Reconstruction: Utilize models from the model zoo.
4. Device Tree Overlay: Configure hardware settings.
5. Application Deployment: Load and run the application code on the development board.

Hardware Design

Connections:

• I2C pins connected to KV260 Pmod pins
• Necessary pull-up resistors enabled for I2C communication

PetaLinux Configuration and Compilation

Steps:

1. Project Creation:

   petalinux-create -t project -s <path_to_bsp_file> --name dpuOS

2. Configuration:

cd dpuOS
petalinux-config --get-hw-description= $PATH_XSA_FILE

1. Enable FPGA Manager
2. Disable TFTPboot Copy
3. *Select EXT4 as Root Filesystem Type

Kernel Configuration

petalinux-config -c kernel

1. Enable DPU Driver
2. Enable USB-to-Serial Converter Driver

Rootfs Configuration

petalinux-config -c rootfs

Build Project

petalinux-build

Packaging and Booting

Prepare SD card and boot the system.

Device Tree Overlay

Generate and compile the device tree overlay to specify hardware configuration.

Running the Application

Install AI Library

1. Download and install Vitis AI Runtime.
2. Optimize DPU for ZynqMP SoCs.
3. Install Vitis AI library and model files.

Run ADAS Application

cd ~/examples/vai_runtime/adas_detection
bash -x build.sh
./adas_detection /dev/video0 /usr/share/vitis_ai_library/models/yolov3_adas_pruned_0_9/yolov3_adas_pruned_0_9.xmodel

DEMO

Demo Video