The latest version of the Raspberry Pi, a small but powerful computer that has turn into quite popular among technology hobbyists and industrial corporations, can now even be an AI computer. On Tuesday, the company published AI kitwhich is a $70 expansion kit with a neural network inference accelerator that will be used for local inference, for the Raspberry Pi 5.
For this latest expansion module, the Raspberry Pi uses the HAT+ expansion card. HAT stands for “Hardware Attached at the Top,” a cute acronym the company uses to designate expansion cards that plug into a regular Raspberry Pi.
The HAT+ expansion card adds an M.2 slot, which is a standard expansion slot commonly used in PC components. For our readers who care about the details: This socket is connected to the Raspberry Pi via a single-lane PCIe 3.0 interface running at 8 Gb/s.
For the kit, the company has partnered with Hailo, an AI chip startup that recently raised $120 million and is trying to challenge Nvidia. Hailo specializes in chips designed to run AI workloads on edge devices reminiscent of cars, smart cameras, robotics and now Raspberry Pi devices.
The accelerator module used by the Raspberry Pi in the AI kit is Hailo-8L. It is a basic module in M.2 format, which suggests it could possibly be easily connected to the HAT+.
Once every thing is installed, you’ll have a Raspberry Pi 5 capable of inferring at a rate of 13 teraoperations per second (TOPS). It’s not much in comparison with an Nvidia GPU, but it’s cost-effective and works with a 27W Raspberry Pi power supply.
On the software side, the latest version of the Raspberry Pi operating system robotically detects the Hailo module so that it could possibly be immediately used by the operating system and applications that use it.
Raspberry Pi has also updated its camera application suite to support neural network inference inside the camera pipeline. For example, it could possibly be used for object detection (“this is a car”), semantic segmentation (“these three things are moving vehicles”), instance segmentation (“these three moving vehicles are a truck, a red car, and a blue car”), pose assessment and face selection.
These are just examples of what you may do with a Raspberry Pi equipped with an AI kit and a first-party or third-party camera. But the Hailo chip will also be used for non-camera applications.
It can be interesting to see how the Raspberry Pi community comes up with latest uses for this kit. This set of AI extensions is a tool, and it’s now as much as Raspberry Pi customers what they wish to do with it.