Nvidia’s Always-On Chip Detects Faces in Less Than a Millisecond

Nvidia researchers have developed an "always-on" chip capable of detecting human faces in under a millisecond. This innovative system significantly reduces power consumption, operating at less than 5 milliwatts while processing video at 60 frames per second. The chip achieves this efficiency by quickly processing data using a deep-learning accelerator and then entering a low-power sleep state, a technique the team calls "race to sleep."

This technology could enable power-saving features in consumer electronics, such as laptops that automatically turn off screens when a user leaves and quickly reactivate upon their return. Potential applications also extend to autonomous vehicles, robotics, and drones, providing continuous visual awareness with minimal energy expenditure.

This advancement could lead to more power-efficient smart devices and autonomous systems that require constant environmental awareness.