Californian companies Maxim Integrated and Aizip have teamed up to reduce the power needed for person detection.

Combining the Max 78000 AI microcontroller and Aizip’s Visual Wake Words (VWW) model brings human-figure detection to IoT image and video at 0.7mJ per inference.

Aizip focuses on artificial intelligence (AI) for applications in the IoT. Maxim Integrated’s Max 78000 neural-network microcontroller detects people in an image using Aizip’s VWW model at just 0.7mJ of energy per inference. This is said to be 100 times lower than using conventional software, and the most economical and efficient IoT person-detection system available.

The low-power network provides longer operation for battery-powered IoT systems that require human-presence detection, including building energy management and smart security cameras.

The mixed precision VWW network is part of the Aizip AIV DNN intelligent vision deep neural network series for image and video applications and was developed with Aizip’s proprietary design automation tools to achieve greater than 85 per cent human-presence accuracy.

The low power consumption allows 13 million inferences from a single AA or LR6 battery.

“The combination of Maxim Integrated’s ultra-low power chip and Aizip’s compact AI models is an important development that will enable many novel and exciting applications in the IoT world,” said Bruno Olshausen at UC Berkeley, an expert in neural computation and neural network models, who also serves as an advisor to Aizip.

Yuan Lu, Aizip’s president, added: “The Max 78000 architecture, toolchain, and example code and models made it easy to get started and hit our accuracy, latency and power targets on schedule.”

Aizip develops AI models for IoT applications. Based in Silicon Valley, it provides design services with performance, fast turnaround time and quick RoI. Its models can be used in applications for an intelligent, automated and connected world.

“Aizip was quick to exploit our per layer quantisation capability to reduce weight storage and achieve a compact, energy-efficient model for human detection,” said Robert Muchsel from Maxim, architect of the Max 78000 microcontroller. “I look forward to working with them on future projects.”