Raspberry Pi 5 as an AI computer for video image analysis and more
The 13 TOPS AI accelerator Hailo 8L is available as a Raspi extension for 80 euros. Even more important is the good integration of drivers and software.

The Raspberry Pi AI Kit and screenshot of the pose estimation application example.
(Image: Hailo)
Rasperry Pi Limited announces the Raspberry Pi AI Kit for the Raspberry Pi 5. It consists of the AI accelerator Hailo 8L, the M.2 adapter M.2 HAT+ and above all the integrated drivers and the development environment in Raspberry Pi OS.
Several examples of (video) image analysis are already available on GitHub, such as object recognition, pose estimation and segmentation. Image analysis with the OpenAI model CLIP (Contrastive Language Image Pre-Training) is soon to follow. The Raspberry Pi AI Kit costs around 80 euros.
Empfohlener redaktioneller Inhalt
Mit Ihrer Zustimmung wird hier ein externer Preisvergleich (heise Preisvergleich) geladen.
Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (heise Preisvergleich) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.
More powerful NPU than in Core Ultra 100 and M1
The Hailo 8L processes up to 13 trillion 8-bit integers (13 TOPS INT8) per second. Its AI performance is therefore higher than that of the Neural Processing Unit (NPU) in the current Intel Core Ultra 100 processors (Meteor Lake, Intel AI Boost). Intel quotes around 11 TOPS for this, but also for more precise INT16 algorithms. The Apple Neural Engine (ANE) of the Apple M1 also achieves 11 TOPS. However, these processors each have even more powerful GPU cores, which can be used for AI algorithms, as well as CPU cores.
The Hailo 8L is primarily designed for image analysis using machine learning (ML).
Advantage: System integration
(Image:Â Hailo)
According to the documentation from Raspi Ltd. and Hailo's GitHub page, commissioning the Hailo 8L under Raspi OS is easy with a simple command - at least much easier than with many other AI accelerators:
sudo apt install hailo-all
Optionally, one of the Raspi cameras can also be integrated, which is typically connected to the Camera Serial Interface (CSI) of the Raspi.
This means that the integration of the NPU in Raspi OS is more advanced than AMD has achieved with the Ryzen AI NPU in Windows 11, for example.
However, the Raspberry Pi AI Kit is not the first AI computing accelerator for Raspis. For several years now, Google's "Coral" accelerators with the Google Edge TPU (4 TOPS) have been connected to Raspis via USB or PCI Express (PCIe/M.2). There was also the Intel Movidius Neural Compute Stick.
Raspi Ltd. has also announced an AI camera module for the Raspi 5, which contains an image sensor with an integrated AI accelerator from manufacturing partner Sony.
(ciw)