The Raspberry Pi AI Kit, developed with Hailo, is now available and costs $70
Raspberry Pi has launched a $70 AI Kit developed with Hailo, allowing the popular single-board computer to integrate high-performance, low-power AI inferencing through a Hailo-8L accelerator on an M.2 HAT+, enabling real-time AI vision applications with extensive software integration and support.
In a context where AI is practically everywhere, it is unsurprising that the Raspberry Pi, everyone's favorite tiny single-board computer, would enter the game with an AI Kit. The $70 AI Kit was developed in collaboration with Hailo. The latter is the edge AI accelerators-focused chipmaker which recently closed a $120 million Series C while looking to take on the big players, NVIDIA included. The AI Kit is made from an M.2 HAT+ (Hardware Attached on Top) preassembled with a Hailo-8L AI accelerator module.
A Raspberry Pi 5 outfitted with an AI Kit can be used to build AI vision applications that run in real-time with low latency and energy consumption. The Hailo-8L runs cutting-edge neural networks for object detection, semantic and instance segmentation, pose estimation, facial landmarking, and other applications, freeing the Raspberry Pi's CPU. The AI Kit HAT+ connects to the Raspberry Pi via an 8 Gbps single-lane PCIe 3.0 connection. The HAT+ adds an M.2 slot, where the AI accelerator connects. With everything set in place, the resulting system is capable of inferencing at up to 13 TOPS —hardly state-of-the-art, but perfectly reasonable considering the module doesn't have energy requirements above the Raspberry Pi’s stock 27W power supply.
The AI Kit also integrates fully with the Raspberry Pi image software subsystem, is compatible with first- and third-party cameras, and can run multiple neural networks on a single camera, or two cameras concurrently with several neural networks or a single one. Hailo's extensive model zoo comprises a variety of deployable pre-trained neural network models optimized to run on the AI Kit. Moreover, the rpicam-apps suite includes a post-processing template that integrates real-time inferencing in the camera pipeline. The pre-installed Hailo Tappas post-processing libraries enable the creation of AI-based applications with a reasonable amount of C++ code. Raspberry Pi plans to release a similar integration for the Picamera2 framework.
In addition to the integrations, a host of resources is available to prospective users: a full set of instructions in the getting started guide, AI demos to try out, Hailo's advanced AI applications running on a Raspberry Pi 5, and a community forum for topics surrounding the Hailo-8L AI accelerator.
