Seeed Studio Aims at Edge AI with the Raspberry Pi 5-Powered reComputer AI R2130-12

Compact chassis hides a Raspberry Pi 5, 26 TOPS Hailo-8 accelerator, and a dual-M.2 board with room for an NVMe SSD too.

Gareth Halfacree
2 months ago β€’ Machine Learning & AI / HW101

Embedded and hobbyist hardware specialist Seeed Studio has opened pre-orders for an all-in-one edge artificial intelligence (AI) machine based around the Raspberry Pi 5 single-board computer and the Hailo-8 machine learning accelerator: the reComputer AI R2130-12.

"reComputer AI R2000 is a powerful Edge AI Computer based on the Raspberry Pi 5. With quad-core Arm Cortex-A76 processor 8GB RAM, M.2 SSD support, and an Hailo-8 AI acceleration module providing up to 26 TOPS [Tera-Operations Per Second], [it enables] real-time, low latency & high-efficiency AI inferencing. it's the perfect choice for a wide range of AI applications, including AI-powered video analytics, machine vision, and intelligent edge computing."

As Seeed says, the compact gadget is based around the Raspberry Pi 5 8GB single-board computer β€” though the likelihood of a follow-up model that takes advantage of the newly-released Raspberry Pi 5 16GB and its doubled memory capacity seems likely. The SBC is housed in a custom chassis, exposing most ports β€” though not the general-purpose input/output (GPIO) header β€” and featuring an active cooling system with a top-mounted fan.

Elsewhere in the case is Seeed's own dual-M.2 add-on board. Unlike Raspberry Pi's own M.2 HAT+, this has two M.2 slots β€” driven from the Raspberry Pi 5's single user-accessible PCI Express lane through an ASMedia ASM2806 PCI Express Gen. 3 switch. One slot features the Hailo-8 M.2 module, featuring a Hailo machine learning accelerator chip delivering 26 tera-operations per second (TOPS) of minimum-precision compute; the other slot is left empty for those wanting to add another accelerator or a Non-Volatile Memory Express (NVMe) solid-state drive β€” though PCIe bandwidth is shared between the two slots.

The company is positioning the kit as ideal for both direct deployment and experimental development, suggesting it could find a use in network video recording (NVR) systems for object detection, analytics, and anomaly detection, in digital signage for facial recognition, smart retail for behavioral tracking, and even in transportation for traffic management.

The reComputer AI R2130-12 is available to pre-order on the Seeed Studio store at $249 before volume discounts, ahead of a planned 10 February launch date.

Gareth Halfacree
Freelance journalist, technical author, hacker, tinkerer, erstwhile sysadmin. For hire: freelance@halfacree.co.uk.
Latest articles
Sponsored articles
Related articles
Latest articles
Read more
Related articles