Overview

Introduction

Hailo-8/8L chip with 26/13Tops computing power

Features

• A neural network inference accelerator built with 13/26 trillion operations per second (TOPS) built on the Hailo-8/8L chip
• High-performance AI expansion board for Raspberry Pi 5
• Installed with hardware kit
• Stacked GPIO headers

Hardware connection

Pay attention to the direction of the cable, and the connection is shown in the figure:

Working with Raspberry Pi

Update

#1. Update the software
sudo apt update && sudo apt full-upgrade
sudo rpi-eeprom-update

#Configure CLI (not required for systems 24 years or later)
sudo raspi-config
#Under Advanced Options> Bootloader Version, select Latest. Then use the Finish or Esc key to exit raspi-config.

#2. Update the firmware
sudo rpi-eeprom-update -a

Identify the device

1. Enable PCIE interface

Connect the hardware, the latest system will have hardware detection, connecting the hardware will automatically enable PCIE
If it defaults to not having the PCIE interface enabled, execute: add "dtparam=pciex1" to /boot/firmware/config.txt

2. Enable PCIE Gen3, then add following to /boot/firmware/config.txt (Gne3 mode must be started):

dtparam=pciex1_gen=3

3. Restart PI5 after modification, and it will then identify the device. (You can choose not to restart immediately, instead waiting to restart after installing the library)

Test the demo

rpicam-apps uses the Hailo AI neural network accelerator to run camera demo

Preparation:

1. Raspberry Pi 5 and Hailo-8 Acce A kit
2. Install 64-bit Raspberry Pi OS Bookworm
3. Install the Raspberry Pi camera (test using Raspberry_Pi_Camera _Module_3 to connect to CAM1 interface)

1. Install the required dependencies for using the AI Kit

sudo apt install hailo-all

2. Restart the device

sudo reboot

3. Check if the driver is working properly

hailortcli fw-control identify

You can also execute dmesg | grep -i hailo to check the log

4. Check the camera

rpicam-hello -t 10s
Please make sure that the camera is working properly

5. Clone the repository rpicam-apps

git clone --depth 1 https://github.com/raspberrypi/rpicam-apps.git ~/rpicam-apps

6. Test

Object detection
rpicam-hello -t 0 --post-process-file ~/rpicam-apps/assets/hailo_yolov6_inference.json --lores-width 640 --lores-height 640
Yolov8 model
rpicam-hello -t 0 --post-process-file ~/rpicam-apps/assets/hailo_yolov8_inference.json --lores-width 640 --lores-height 640
YoloX model
rpicam-hello -t 0 --post-process-file ~/rpicam-apps/assets/hailo_yolox_inference.json --lores-width 640 --lores-height 640
Yolov5 character and face model
rpicam-hello -t 0 --post-process-file ~/rpicam-apps/assets/hailo_yolov5_personface.json --lores-width 640 --lores-height 640
Image segmentation
rpicam-hello -t 0 --post-process-file ~/rpicam-apps/assets/hailo_yolov5_segmentation.json --lores-width 640 --lores-height 640 --framerate 20
Pose estimation
rpicam-hello -t 0 --post-process-file ~/rpicam-apps/assets/hailo_yolov8_pose.json --lores-width 640 --lores-height 640

If the above command prompts "ERROR: *** No such node (hef_file) ***" , execute the following

Object detection
rpicam-hello -t 0 --post-process-file /usr/share/rpi-camera-assets/hailo_yolov6_inference.json --lores-width 640 --lores-height 640
Yolov8 model
rpicam-hello -t 0 --post-process-file /usr/share/rpi-camera-assets/hailo_yolov8_inference.json --lores-width 640 --lores-height 640
YoloX model
rpicam-hello -t 0 --post-process-file /usr/share/rpi-camera-assets/hailo_yolox_inference.json --lores-width 640 --lores-height 640
Yolov5 character and face model
rpicam-hello -t 0 --post-process-file /usr/share/rpi-camera-assets/hailo_yolov5_personface.json --lores-width 640 --lores-height 640
Image segmentation
rpicam-hello -t 0 --post-process-file /usr/share/rpi-camera-assets/hailo_yolov5_segmentation.json --lores-width 640 --lores-height 640 --framerate 20
Pose estimation
rpicam-hello -t 0 --post-process-file /usr/share/rpi-camera-assets/hailo_yolov8_pose.json --lores-width 640 --lores-height 640

For more information, please visit GitHub, Hailo official website