Raspberry Pi AI Kit hands-on: An affordable way to enhance the SBC’s performance in AI tasks

The Raspberry Pi ecosystem is full of accessories, HATs, and sensors that can boost your SBC model’s capabilities to the next level. With the Raspberry Pi 5 adding a PCIe interface to its already robust port selection, you can connect even more attachments to the SBC. One of these is the recently released Raspberry Pi AI Kit, which consists of an M.2 HAT+ and a Hailo-8L NPU.

If AI modules for the Raspberry Pi 5 sound familiar, that’s because Coral released a $60 USB Accelerator a while ago. But what’s unique about the official first-party AI Kit is that it adds a whopping 13 TOPS of processing capabilities to the RPi 5 compared to the 4 TOPS provided by Coral Kit.

With such insane numbers, you might be wondering if there’s a catch to the accessory. As it turns out, the biggest issue with the Raspberry Pi AI Kit is that it’s geared specifically toward developers rather than the average user who just wants to experiment with LLMs. Don’t get me wrong: In the right hands, this inexpensive accessory can be a real game-changer. But for the mainstream audience, it’s easy to get bored with the device after using it to run a handful of object detection demos.

Raspberry Pi AI Kit

The Raspberry Pi AI Kit bundles an M.2 HAT+ with a Hailo AI acceleration module, allowing you to use the NPU with the Raspberry Pi 5. It provides an accessible, cost-effective, and power-efficient way to integrate high-performance AI.

About this article: I purchased the Raspberry Pi AI Kit last month to satisfy the tinkerer residing in my brain. So, the Raspberry Pi Foundation had zero input into the contents of this article.

Pricing and availability

The Raspberry Pi Foundation released the AI Kit at the beginning of June, with the device retailing for $70. But as is the fate of every first-party accessory for the uber-popular lineup, you’ll have a hard time finding the Raspberry Pi AI Kit due to its low stock. As of writing, plenty of retailers, including CanaKit, CED, and Vilros, have it listed on their websites, though you might need to wait a while, as it’s sold out on most websites. Well, you can technically grab it from Amazon, but you’ll have to shell out almost twice as much as the device’s launch price.

Setup and installation

Relatively painless for the most part

The Raspberry Pi AI Kit includes everything you need to mount it on top of the RPi 5, and the overall process is pretty straightforward. First, you’ll have to install the four standoffs on the base of the Raspberry Pi 5 and place the GPIO extender on top of the pins. Once that’s done, you can plug the ribbon cable on the M.2 HAT+ of the AI Kit into the PCIe socket of the Raspberry Pi 5 before using the last set of screws to secure the accessory on top of the standoffs. Just don’t forget to connect a camera module with the SBC afterward.

Configuring the software aspect is just as simple, though you’ll need to be on the 64-bit version of the Raspberry Pi OS to use the camera. All you have to do is enable PCIe 3.0 support in the raspi-config settings and run the sudo apt install hailo-all command to install the necessary dependencies for the AI Kit.

Software

A bit lacking for the average consumer

When I initially heard about the AI Kit, I was dying to test it with some LLMs and compile the benchmarks for most of the popular models. When I tried my hand at running some language models on my RPi 5 via Ollama a few months ago, the results were far from impressive. But as it turned out, the AI Kit isn’t powerful enough to drive full-fledged LLMs.

Instead, the Hailo-8L processor is only compatible with machine-learning tasks involving the feed captured by camera modules. And yes, I meant camera modules, not just any old webcam or IP camera. That’s because the demos available on GitHub only support the RPiCam package, which, in turn, runs on camera modules. Having tested both my OBSBOT Tail Air and Aicoco onAir webcams in different modes, I can confirm that you’ll need a proper camera module that connects to the MIPI socket.

Nevertheless, I fired up the Object Detection demo using the Yolov6 interface, and the performance was pretty decent for the most part. That said, there’s still a lot to be done to improve the accuracy of the AI Kit. In many cases, the Raspberry Pi failed to detect all objects in the frame. Even when it did display a boundary around an item, it wasn’t able to identify said object properly.

Switching to Yolov8 improved things quite a bit. While it wasn’t perfect by any means, Yolov8 was significantly better at identifying the objects in the frame. I also tested the YoloX model, and its performance and identification capabilities were somewhere in between those of Yolov8 and Yolov5. That said, despite the issues with Yolov5 in the Object identification, the model performed surprisingly well in the Person + Face detection test.

Image Segmentation test using Yolov5

Next, I ran the Image Segmentation test. Even after repeating the test with different objects, the AI Kit could only color one item in the entire frame, so it wasn’t as successful as the other tests.

Meanwhile, the Pose Estimation test produced the best results so far, as the Raspberry Pi was able to create a surprisingly accurate wireframe version of my posture.

Apart from these tests, the Hailo-8L NPU is compatible with a handful of other AI models that you can acquire from the official GitHub page. While it’s good to know that the NPU will integrate support for picamera2 and CLIP apps, that’s pretty much all you can do with the Kit at the moment.

Should you buy the Raspberry Pi AI Kit?

The AI Kit is worth your money if:

• You’re a developer who needs better performance in object and pose recognition tasks.
• You don’t mind paying $70 for an AI Kit with a limited number of demos.

You should steer clear of the AI Kit if:

• You want more horsepower when running LLMs on your Raspberry Pi.
• You’re not a big fan of artificial intelligence projects and just want to bring some cool non-AI/ML ideas to life with your SBC.

From the title, you can tell that this wasn’t going to be a review where I dissect every aspect of the AI Kit – as there’s not a lot I could experiment with besides a handful of demos. Personally, I consider the AI Kit to be worth the $70 I spent. But that’s because I have some project ideas that can benefit from the Hailo-8L’s 13 TOPS of performance. Sure, it may not be able to run LLMs, but the superior pose detection and face identification capabilities of the AI Kit can help out with automation projects.

But that’s just my perspective as someone who’s into coding. If you’re unwilling to write Python codes or take the plunge into the massive world of Pytorch and other machine-learning tools, the AI Kit isn’t for you. Therefore, I’d advise the average user to check out useful accessories instead of spending their money on the AI Kit.

Related

10 best operating systems for Raspberry Pi 5

Unsure about which OS you should use with your Raspberry Pi 5? Here are ten operating systems worth running on the palm-sized system