Maxim Integrated and Aizip, a company focused on artificial intelligence (AI) for applications in the Internet of Things (IoT), have declared that Maxim Integrated’s MAX78000 neural-network microcontroller can detects people in an image using Aizip’s Visual Wake Words (VWW) model at just 0.7 millijoules (mJ) per inference.
This is 100 times lower than conventional software solutions, and the most economical and efficient IoT person-detection solution available.
“The combination of Maxim Integrated’s ultra-low-power chip solutions 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, a highly recognized expert in neural computation/neural network models who also serves as an advisor to Aizip.
“The MAX78000 architecture, toolchain, and example code and models made it easy to get started and hit our accuracy, latency and power targets on schedule,” said Yuan Lu, Co-Founder and President, Aizip.
“Aizip was quick to exploit our per layer quantization capability to reduce weight storage and achieve a compact, energy-efficient model for human detection. I look forward to working with them on future projects,” said Robert Muchsel, Maxim Integrated Fellow and architect of the MAX78000 microcontroller.
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 MAX78000 low-power, the neural-network accelerated microcontroller executes AI inferences at less than 1/100th of the energy of conventional software solutions to dramatically improve run-time for battery-powered edge AI applications.
The mixed precision VWW network is part of the Aizip Intelligent Vision Deep Neural Network (AIV DNN) series for image and video applications and was developed with Aizip’s proprietary design automation tools to achieve greater than 85 percent human-presence accuracy.
- Extended Battery Life: Efficient AI model and low power microcontroller system-on-chip (SoC) reduce inference energy to 0.7 mJ, allowing 13 million inferences from a single AA/LR6 battery.
- Cost-Effective Intelligence at the Edge: Extreme model compression enables accurate smart vision with a memory-constrained, low-cost AI accelerated microcontroller and budget-friendly image sensors.