Scientists from the University of Washington and Princeton University have developed a groundbreaking compact camera designed for computer vision.

 

 

Scientists from the University of Washington and Princeton University have developed a groundbreaking compact camera designed for computer vision. This innovative prototype uses optics for computing, slashing power consumption while enabling object recognition at the speed of light.

 

Instead of traditional glass or plastic lenses, the camera relies on 50 layers of meta-lenses—flat, lightweight optical components with microscopic nanostructures that manipulate light. These meta-lenses function as an optical neural network, performing calculations instantly before the image even reaches the sensor.

 

By shifting computation from electronics to optics, the camera identifies and classifies images over 200 times faster than neural networks running on traditional computer hardware—while maintaining comparable accuracy. Since it operates on incoming light instead of electricity, it also dramatically reduces power consumption.

 

This breakthrough could transform AI-powered devices, from self-driving cars and autonomous robots to medical devices and smartphones. Researchers plan to refine the prototype to handle complex object detection—a key function for autonomous navigation.

Published in Science Advances, the study brings deep learning into optics, demonstrating a new era of computer vision technology. The researchers aim to push further, developing systems capable of tackling more demanding AI applications.