Raspberry Pi Teams Up with Hailo to Enhance AI Capabilities with New Extension Kit

The latest Raspberry Pi version, a compact yet powerful computer that's gained traction among tech enthusiasts and industrial firms, is now designed for AI applications. On Tuesday, the company unveiled the AI Kit—an affordable $70 extension featuring a neural network inference accelerator tailored for local inferencing on the Raspberry Pi 5.

This new extension leverages the HAT+ (Hardware Attached on Top) system, a clever term the company uses for its extensions that securely connect to the Raspberry Pi.

The HAT+ card introduces an M.2 slot, a standardized connector often utilized in PC hardware. For tech aficionados interested in specifics, this slot links to the Raspberry Pi through a single-lane PCIe 3.0 interface operating at 8Gbps.

In partnership with Hailo—a pioneering AI chip startup that recently secured $120 million funding to compete with Nvidia—Raspberry Pi is embracing cutting-edge AI workloads. Hailo focuses on developing chips optimized for edge device AI tasks, including applications in vehicles, smart cameras, robotics, and now Raspberry Pi units.

The AI Kit features the Hailo-8L, a user-friendly entry-level accelerator in an M.2 form factor, which conveniently plugs into the HAT+. Once installed, your Raspberry Pi 5 can perform inferencing at an impressive 13 tera-operations per second (TOPS). While this speed may not rival Nvidia GPUs, it remains a budget-friendly solution compatible with the Raspberry Pi’s standard 27W power supply.

On the software side, the latest Raspberry Pi OS automatically recognizes the Hailo module, enabling immediate compatibility with applications designed to utilize its capabilities.

Raspberry Pi has also enhanced its suite of camera applications to support neural network inferencing seamlessly within the camera pipeline. Users can implement 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 estimation, and facial landmarking.

These examples illustrate the potential of a Raspberry Pi equipped with the AI Kit and either a first-party or third-party camera. However, the versatility of the Hailo chip extends beyond photography, opening up numerous non-camera use cases.

It will be exciting to see how the Raspberry Pi community innovatively utilizes this kit. The AI extension kit is a powerful resource, and the creative possibilities now lie in the hands of Raspberry Pi users.