Nuvoton Targets Visual Gesture Recognition at the Edge with the NuMaker-GestureAI-M55M1

Compact computer vision board aims to deliver on-device gesture recognition and person location at a low power draw.

Gareth Halfacree
14 seconds agoAI & Machine Learning / HW101

Nuvoton has announced a new development board, built around its MM55M1 microcontroller, designed for low-power on-device computer vision for gesture control at the edge: the NuMaker-GestureAI-MM55M1.

"Engineered to help developers overcome technical barriers in AI [Artificial Intelligence] model deployment and significantly reduce R&D [Research & Development] timelines," Nuvoton claims of the new board, "this 'ready-to-integrate' platform is powered by the NuMicro M55M1 series microcontroller. By providing a robust hardware environment, the module empowers developers to rapidly implement intelligent gesture recognition and visual perception into smart home appliances, industrial equipment, and medical devices. The NuMaker-GestureAI-M55M1 effectively accelerates the journey from prototyping to mass production."

While Nuvoton is positioning the board as riding high on the current interest in artificial intelligence, there's no large language model (LLM) involved in the new board. Instead, it's a pretty traditional computer vision approach, based on a model that runs entirely locally on the MM5M1's Arm Ethos-U55 neural coprocessor. Alongside this, the chip also includes a single 32-bit Arm Cortex-M55 core running at up to 220MHz, Helium accelerator for digital signal processing, up to 1.5MB of static RAM (SRAM), and up to 2MB of flash storage.

The NuMaker-GestureAI-M55M1 deploys the chip in a board alongside a GalaxyCore GC0308 VGA (640×480) CMOS image sensor. A microSD slot provides expanded storage for model files, with a pre-loaded model capable of recognizing "more than 10" gestures including "Call," "Like," "OK," and "Stop," along with person-recognition and location-reporting capabilities. The board is also compatible with the Nuvoton AI toolchain, allowing users to deploy custom models, including TensorFlow Lite Micro models, onto the board.

More information is available on the NuMaker-GestureAI-M55M1 product page, with the board listed on the Nuvoton Direct store at $30 — though, at the time of writing, it was showing as out of stock.

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