A Raspberry Pi 4, top, and Pi 5, bottom.

A quick play with the AI Kit by Raspberry Pi

Reading Time: 2 minutes

Yesterday afternoon as I set off for my walk I noticed that the Pi AI Kit was in my post box. I moved it to the locked compartment and then went for my walk. After the walk I took the time to set it up and it went well.

## Easy Assembly

If you’ve got your own kit but haven’t assembled it first, the long screws are there so that you can fix the Raspberry Pi and AI kit within the default Pi case. I made the mistake of using the shorter screws and had to swap them. It delayed me only slightly.

## Affordable

The advantage of the [Raspberry Pi AI Kit](https://www.pi-shop.ch/raspberry-pi-ai-kit) is that it’s 77.90 CHF at the point of writing this blog post. Nvidia solutions, from a quick glance start at 180 CHF. If you want to play with Apple Intelligence you will need to spend either almost a thousand francs on a new phone or 1000 CHF or more for a Mac Book air, and more for “better”. In other words the Pi is affordable for those who want to experiment.

My experience of installing the hardware as well as the firmware was easy. I was able to download the git repository and run a test with the detection.py file. I tested their footage as well as my own and it went smoothly.

## Experimentation

I used the [Hailo Raspberry Pi 5 Examples](https://github.com/hailo-ai/hailo-rpi5-examples) for my tests. When I ran my own footage rather than the default footage it detected the bench, a stop sign that is an emergency floatation device, and one or two people when waves broke in a certain manner.

## Project Ideas

If you’re looking for [AI Kit project ideas](https://www.jaycon.com/10-fun-projects-for-the-new-raspberry-pi-ai-kit/).

## Cooling Concern

The Pi default case and AI Kit Hat fit well, but I worry that the AI Kit is prone to overheating. It is currently too hot to touch for more than a few seconds. I also worry that by placing the hat I have reduced air flow to the Pi.

## And Finally

We are at the beginnng of the adoption curve so it will become more useful with time. I expect that we will find more guides and projects in the coming weeks and months. For now I was able to detect objects in video, but can’t tell whether it’s being used by Photoprism, Immich or Nextcloud. That will come later.