Have you ever thought of getting a small camera that can surpass any other device but uses little power? Well, a group of scientists has been able to achieve this with the help of a new compact camera that simply needs lights to complete these tasks. This research was completed by two scientists and their team of talented students. What they achieved is a device capable of identifying objects at the speed of light. Here’s what the camera is all about.
The lead image is a screenshot from the UW Department of Electrical & Computer Engineering.
According to Science Advances, Arka Majumdar, a professor of electrical and computer engineering and physics at the University of Washington, and Felix Heide, an assistant professor of computer science at Princeton University, worked along with their students to create the device. What makes the compact camera different is that it has special, tiny 50 lenses called “metalenses,” which use light to process the image before the camera captures it. In other words, it performs large-kernel spatially varying (LKSV) convolutions before the sensor records the data. The metalenses here work as an optical neural network, which is basically an AI computer system designed to work as a human brain. It is a great deal that the camera can recognize objects and patterns using light without the use of any electricity. To test this theory, the scientists used a smartphone display and a large-area sensor to capture photographs.
In addition, this compact camera is said to be super thin, about the thickness of a few paper sheets. But its size will not hinder its work. In addition, the device uses over 99% less power than traditional cameras because it relies on light instead of electricity for most of its computations.
On the whole, the most impressive thing is the compact camera’s accuracy. When tested on CIFAR-10, a standard set of photographs, the camera could recognize objects with 72.76% accuracy, beating other traditional computer programs like AlexNet. Furthermore, one can use it for complex tasks like semantic segmentation, where the device can label different parts of the photographs after identifying them. For instance, being able to differentiate between the sky, buildings, and people in an image.
At the moment, this device can be helpful in medical imaging, for instance, to identify different parts of an X-ray or MRI scan. It can also be utilized in transportation to help with innovations such as self-driving cars, as it will be able to recognize objects on the road more easily. It can also be used for technology such as smartphones, smart glasses, or even small robots.
This technology, while in its infancy, can really help photographers, too, have a camera that can survive longer without needing to be charged repeatedly. At the same time, when perfected, this will also help with the challenges of tracking issues in cameras. Other than that, if the technology is first used in smartphones, the chances of it making them better will be come higher.
However, it remains to be seen how this compact camera can be leveraged by camera manufacturers for their lineup. Perhaps, it’s potential is utilized for the better or it is just another experiment that remains restricted to the field of science.