Raspberry Pi AI Kit - Unboxing and Installation Guide

Science & Technology

Introduction

Hey everyone, I'm Gilad and I lead the makers and developers community at Halo. Today, I am very excited to introduce Raspberry Pi's new AI kit featuring the Halo 8l entry-level AI accelerator. This kit delivers 13 TOPS with a typical power consumption of around 2 Watts. You can find the kit starting today at official Raspberry Pi resellers.

Along with this release, we are also launching the Halo Community platform and opening our developer zone to all users. You can find the links in the description. We have put a lot of effort into ensuring that the installation flow and examples are as straightforward as possible, enabling you to keep data processing local, ensuring your privacy, optimizing performance, and managing costs according to your preferences. All of Halo's examples are open source, and we encourage you to use them in your projects and products.

Today, we are releasing three basic pipelines for different tasks: detection, pose estimation, and instant segmentation. These pipelines are built in Python for easy integration. Additionally, Raspberry Pi has integrated Halo inference into its official rpy cam apps repo, which is Raspberry Pi's C++ camera framework.

In this video, we will review the Halo installation flow on the Pi, introduce our new GitHub repository, and explore the available examples. Let's jump right into it!

Unboxing and Setup

To begin, you will need the following:

  • Raspberry Pi 5
  • Raspberry Pi AI Kit
  • Micro HDMI to HDMI adapter
  • Active cooler
  • Raspberry Pi Camera Model 3 or High-Quality Camera
  • Raspberry Pi Display Cable
  • 27 USB-C Power Supply

Start by unboxing the Raspberry Pi 5 and connect the active cooler. Make sure to connect the fan to its connector. Prepare the Raspberry Pi camera and connect it. Finally, open the AI kit, which comes pre-installed with a thermal pad between the MD2 and the board. If your design allows for sufficient ventilation, you may not need an additional heat sink.

Once everything is connected, you are ready to go!

Installation Guide

  1. Update the system by running sudo apt-get update and sudo apt-get upgrade.
  2. Set the PCI speed to Gen 3 for optimal performance. Use the Raspberry Pi raspi-config tool to enable this setting.
  3. Reboot the Pi.
  4. Install Halo software by running sudo apt install halo.
  5. Reboot the Pi once again.

To verify the installation, run the following commands:

  • halo-rt-cli firmware control identify to ensure the chip is identified correctly.
  • halo-rt-cli elements list to check the installed Halo elements.

If everything works fine, proceed to the Halo RPi examples and demos.

Halo RPi Examples

Visit the Halo RPi GitHub repository to access the Halo examples. There, you will find the installation guide, examples for detection, pose estimation, and instant segmentation, as well as additional documentation.

To run the examples, follow these steps:

  1. Clone the RPi Basic Pipelines repository.
  2. Configure your environment by sourcing the setup.sh script.
  3. Install the requirements specified in the requirements.txt file.
  4. Run the example scripts provided.

Each example demonstrates how to utilize the Halo AI kit for different tasks. For the detection example, you can choose between three different networks: YOLO V6n, YOLO V8s, and YOLO XS leaky. The pose estimation example detects persons and prints their left and right eye coordinates. In the instant segmentation example, detections are printed to the terminal while running on a video file.

Additionally, the examples provide various options, such as controlling the input source (file, USB camera, or Raspberry Pi camera), displaying FPS, enabling additional post-processing, and selecting different networks.

Remember to visit the Halo Community forum for support and discussions related to the Halo AI kit.

Keywords:

Raspberry Pi AI Kit, Halo 8l AI accelerator, installation guide, Python integration, examples, detection, pose estimation, instant segmentation, Raspberry Pi resellers, Halo Community platform

FAQ:

Q: Where can I find the Raspberry Pi AI Kit? A: The kit is available at official Raspberry Pi resellers.

Q: Are the Halo examples open source? A: Yes, all Halo examples are open source.

Q: What are the three basic pipelines released with the Raspberry Pi AI Kit? A: The three basic pipelines are for detection, pose estimation, and instant segmentation.

Q: How can I control the input source for the examples? A: You can choose between file, USB camera, or Raspberry Pi camera as the input source.

Q: Can I run the Halo AI Kit on Raspberry Pi 5? A: Yes, the Halo AI Kit is compatible with Raspberry Pi 5.

Q: How do I configure the Halo AI Kit for optimal performance? A: Set the PCI speed to Gen 3 using the raspi-config tool.

Q: Are there any additional resources available? A: Yes, you can find more documentation and join the Halo Community forum for support and discussions.