Maxim Integrated Products' MAX78000 Ultra-low power neural network microcontroller uses Xailient's proprietary Detectum neural network to detect and locate faces in video and images. Xailient's neural network consumes 250 times less power than traditional embedded solutions (only 280 microjoules). Every 12 milliseconds of reasoning, the network performs in real time, faster than the most efficient edge face detection solutions.
These apps use a low-power "listen" mode and then wake up a more complex system when a face is detected. The company's microcontrollers are paired with Xailient's neural network to improve the overall power efficiency and battery life of the edge/cloud hybrid application.
Xailient's Detectum neural network, which includes focus, zoom and visual wake word technologies, detects and locates faces in video and images 76 times faster and with similar or higher accuracy than traditional software solutions. In addition, flexible networks can be extended to applications beyond facial recognition, such as livestock inventory and monitoring, parking space occupancy, and stock water equality.
"Xailient Detectum Neural network, the MAX78000 is able to classify and localize, in addition to seeing face images or videos, you can also determine the field of view of those face images," says Robert Muchsel of Maxim MAX78000 MICROcontroller integration and architect. "Advanced applications include counting people, vehicles and objects, presence or obstacle detection, as well as path mapping and pedestrian heat mapping."
"Ai is poised to become the second largest carbon emitting industry," said Dr. Shivy Yohanandan, CTO of Xailient and inventor of Xailient Detectum Neural network technology. "Replacing 14 traditional Internet protocol cameras using traditional cloud AI with edge-based cameras equipped with Maxim Integrated MAX78000 and Xailient neural network has the same carbon impact as taking a gasoline-powered car off the road."
