False triggering of camera systems that need to wake-up when a person is present is more than a nuisance, it unnecessarily consumes energy, digital storage space and communications bandwidth. Plus, capturing photos or video of things you don't care about and sending cell phone alerts for no reason not only wastes resources, it also wastes your time. A smart wake-up solution from Maxim Integrated and Aizip can drastically reduce false positives to ensure the system stays off until really needed.
The Aizip Visual Wake Words (VWW) or person detector solution running on the Maxim Integrated MAX78000 AI Microcontroller detects one or more people in the field of view with a minimum of power. The solution is truly smart as the Aizip deep neural network recognizes human figures in the field of view of a low-cost camera. The network inference executes in 18ms and consumes just 0.7mJ of energy on the MAX78000. The complex model is trained on Aizip's state of the art pipeline and uses mixed-precision weights to maximize pattern recognition with a minimum of memory resources.
Smart wake-up is also valuable for interactive touchscreens, kiosks, door entry controllers, and anything that needs to be active when a person is present. Smart human detection can increase safety by alerting the presence of a person near vehicles, robots, or other equipment. The equipment can halt operation to protect the person or itself.