Ohm Lab's Neuro N6 Delivers Milliwatt-Scale Computer Vision at the Edge — with Arduino Compatibility
Designed for Arduino developers looking to make the jump into edge vision, the Neuro N6 is built around STMicro's STM32N6.
Internet of Things startup Ohm Lab has launched a crowdfunding campaign for a Feather-format Arduino-compatible STMicroelectronics STM32N6-based development board built with low-power computer vision projects in mind: the Neuro N6.
"Edge AI [Artificial Intelligence] has enormous potential, but getting started is often difficult. Developers are faced with complex toolchains, fragmented ecosystems, and the need for specialized expertise," claims Ohm Lab's Mukund Srinivasa Raghavan. "This is a challenge we have experienced first hand, and it is exactly what Neuro N6 is designed to solve. Neuro N6 is a compact edge AI development board that brings advanced machine learning directly to your hardware projects."
The board itself follows the familiar breadboard-friendly Feather footprint, with its general-purpose input/output (GPIO) pins brought out to a pair of 0.1"-spaced headers on its long edges. Powering the board is an STMicro STM32N6 microcontroller, though it's not the 800MHz Arm Cortex-M55 core that's the main focus of the device; instead, Ohm Lab is focusing on what you can do with the on-board coprocessor, a 1GHz Neural-ART neural processing unit (NPU) delivering 600 giga-operations per second (GOPS) at a milliwatt power draw. Elsewhere there's 64MB of pseduo-static RAM (PSRAM), 342MB of flash, an on-board microphone, and six-degrees of freedom inertial measurement unit with magnetic compass, plus a 500mA lithium-polymer battery charging circuit.
"Neuro N6 is one of the first microcontroller based development boards with a dedicated neural processing unit that is compatible with the Arduino ecosystem," Raghavan claims. "For most users, this means you can implement the entire intelligent system within a single Arduino sketch, moving seamlessly from sensing, to AI inference, to real world action."
Perhaps unusually for a beginner-friendly computer vision development board, the Neuro N6 doesn't include an on-board image sensor. Instead, its creators have opted to add a board-to-board connector for three different types of image sensor: an Omnivision OV5640 five megapixel rolling-shutter camera and time-of-flight sensor; the OV5450-W, which adds Wi-Fi and Bluetooth Low Energy (BLE) connectivity; a FLIR Lepton 3.1R 160×120 thermal camera module; and an adapter board for any STMicro camera module, delivered with the 1.5-megapixel global-shutter VD66GT sensor as standard.
"Alongside the hardware platform, Ohm Lab has developed NeuroStudio, a lightweight, open source, and cross platform application designed to streamline development and debugging on the Neuro N6," Raghavan adds. "NeuroStudio allows you to write your code in Arduino and instantly view the live camera feed and inference results directly over USB, without the need to configure USB webcam drivers or external tools."
The Neuro N6 is currently funding on Kickstarter, with physical rewards starting at £54 (around $72) for "early bird" backers of a single Neuro N6, without camera board; the cheapest complete bundle is priced at £89 (around $119) with the Omnivision OV6540 camera board. All hardware is expected to ship in November this year, but as with all crowdfunding campaigns fulfillment is not guaranteed.