Maxim Integrated, Aizip claim lowest power IoT person-detection solution

Maxim Integrated Products, Inc. and Aizip Inc. have partnered to provide a low-power IoT person-detection solution. Maxim Integrated’s MAX78000 neural-network microcontroller using Aizip’s Visual Wake Words (VWW) model detects people in an image at 0.7 millijoules (mJ) per inference, claiming it is 100× lower than conventional software solutions. This translates into 13 million inferences from a single AA/LR6 battery.

The low-power MAX78000 neural-network accelerated microcontroller (MCU) executes AI inferences at less than 1/100th the energy of conventional software solutions, which improves run-time for battery-powered edge AI applications, said Maxim Integrated. It offers a combination of a dedicated neural network accelerator with two microcontroller cores that enables machines to see and hear complex patterns with local, low-power AI processing that executes in real-time. This means much higher power efficiency for applications such as machine vision, audio, and facial recognition.

The architecture of the MAX78000 (Image: Maxim Integrated)

The mixed precision VWW network is part of the Aizip Intelligent Vision Deep Neural Network (AIV DNN) series for image and video applications. Developed with Aizip’s proprietary design automation tools, the VWW model achieves greater than 85 percent human-presence accuracy, according to the company.

The automated design tools take into account the design targets and hardware parameters such as processing power and memory size, said Hanlin Sun, test engineer at Aizip, in a demo video. He said quantization was part of the design process that reduces weight of the model, which requires minimum processing power of memory while maintaining a high accuracy at a low latency.

The VWW model was developed specifically for IoT applications. Together, the MAX78000 with the AWW model targets battery-powered IoT systems that require human-presence detection, including building energy management and smart security cameras. Click here for a video on how the IoT person-detection solution works. The video shows both indoor and outdoor demos using the AIV VWW Tiny ML model running on the MAX78000 chip.

The VWW detects the presence of people and sends the wake signal to the control system and it can be used in many applications such as building lights control, beach safety monitoring, and home surveillance, said Sun.

The MAX78000 evaluation kit includes a digital microphone, a gyroscope/accelerometer, parallel camera module support, and a 3.5-inch touch-enabled color TFT display. It also features a USB to SPI bridge that provides access to onboard memory, allowing large networks or images to load quickly.

The MAX78000 evaluation kit (Image: Maxim Integrated)

 The MAX78000 microcontroller and MAX78000EVKIT# evaluation kit are available now at Maxim Integrated’s website and through authorized distributors. The MAX78000 is priced at $8.50 (1000-up, FOB USA) and the evaluation kit is $168.00. AIV DNN series models, tools and services are available directly from Aizip.