Raspberry Pi’s New Entry-Level AI Kit Leverages Hailo Accelerators

Raspberry Pi has selected Hailo, a company that develops edge artificial intelligence (AI) processors, to supply AI accelerators for the Raspberry Pi AI kit. The add-on is designed to work with the latest Raspberry Pi 5. This collaboration will allow professional and enthusiast creators to add advanced AI capabilities to their projects.

 

What Comes In the Raspberry Pi AI Kit?

The Raspberry Pi AI kit includes the latest Raspberry PI M.2 HAT+ preinstalled with the Hailo-8L M.2 AI acceleration module for 13 tera-operations per second (TOPS) inference capability. The M.2 HAT+ communicates between the AI accelerator's M.2 interface and the Raspberry Pi 5's PCIe 2.0 interface using a single-lane PCIe 3.0 connection running at 8 GBps for seamless data exchange.

The kit (datasheet linked) fully integrates with Raspberry Pi's camera software stack and supports third-party camera modules, thanks to Hailo's software suite and Model Zoo. The accompanying software, including the rpicam-apps and the upcoming Picamera2 framework integration, allows developers to create advanced AI applications with only a few hundred lines of C++ code. The kit also supports inferencing on pre-recorded videos through the GStreamer framework. Its concurrent processing capability can run multiple neural networks parallelly on one or two cameras.

 

RPi AI kit with RPi 5. Image used courtesy of Raspberry Pi

 

The Raspberry Pi AI kit comes with detailed documentation to help users install the required software packages and run demo applications. Further support is available on the Hailo community forum, including tutorials, FAQs, and other resources to help creators in their projects. The community forum allows registered members to connect with a team of Hailo experts and each other to share codes, experiences, and materials around their projects.

 

Hailo AI Accelerator Supports 13 TOPS

Unlike cloud-based solutions, edge AI modules improve the computing capacity and effectiveness of AI processing directly on devices. This approach offers several advantages. It uses less bandwidth and computing power by transferring less data. It also enables real-time decision-making, which is critical for applications like autonomous robots or industrial automation systems.

Incorporating AI capabilities into prototyping solutions like Raspberry Pi significantly benefits professional developers and hobbyists. Professionals can rapidly test AI-driven applications on a versatile, cost-effective platform to accelerate product development. Hobbyists, on the other hand, gain access to powerful AI tools that were previously limited to high-end systems. With edge AI, the possibilities for creative exploration in various fields like robotics and home automation become much richer.

The Hailo-8L AI accelerator, included in the Raspberry Pi AI kit, features up to 13 tera-operations per second (TOPS), enabling real-time, low-latency, and high-efficiency AI inferencing on edge devices.

 

The Hailo-8L is an entry-level AI accelerator. Image used courtesy of Hailo 

 

It can concurrently process AI tasks with multiple models and offers high power efficiency, with a typical power consumption of 1.5 W. This makes it suitable for solutions requiring limited AI capacity and low power consumption. In addition, Hailo claims its accelerator is scalable for products in the prototype or mass production stage.