Smart glass is trained similar to machine learning, "learning" to bend light to the right place. Pictures from the Internet
According to the latest issue of "Photonics Research" magazine, American researchers creatively designed a "smart" glass that can recognize and distinguish images using light only, and can operate without electricity or sensors.
A team led by Professor Yu Zongfu from the Department of Electrical and Computer Engineering at the University of Wisconsin-Madison used optical systems to compress the normal settings of cameras, sensors and deep neural networks into a thin glass. In the proof-of-concept study, this smart glass can distinguish handwritten unit digits, and its dynamic is enough to detect the handwritten 3 being rewritten to 8 in real time.
The working principle of smart glass is to embed bubbles of different sizes and shapes, as well as small pieces of light-absorbing material such as graphene into specific locations of the glass. When the light waves reflected by the handwritten digits pass through the glass, they are bent by bubbles and impurities in a specific way. This bending process focuses the light to one of 10 specific points on the other side of the glass, depending on the amount presented. The 10 points correspond to the numbers 0 to 9, respectively. Yu Zongfu described it as a key to open a lock. Even when writing in different styles, the glass can focus light to detect the correct number.
Designing glass to recognize numbers is similar to the training process of machine learning, except that the training object is analog materials rather than digital codes. If the light is not focused at the correct point through the glass, the research team will adjust the size and position of the impurities slightly. After thousands of iterations, the glass "learned" to bend the light to the right place. Yu Zongfu said that these impurities are basically like "artificial neurons."
Yu Zongfu pointed out that the wave dynamics of light propagation provide a new method for simulating artificial neural computing. The advantage is that the calculation is completely passive and inherent in the material, which means that smart glass that can recognize images can be used hundreds of thousands of times.
At present, when people use facial recognition to unlock their phones, artificial intelligence takes up a lot of computing resources and consumes electricity. In the future, smart glass may be used as a "biometric lock". A piece of glass only recognizes a person's face, thereby greatly improving the security of the mobile phone. This small, inexpensive, one-time smart product will be able to operate under extremely tight energy, bandwidth, and financial resources.
However, the practicality of smart glass remains to be further studied, mainly because smart glass still lacks the flexibility to run on a computer processor and can be used for many different tasks. (Reporter Feng Weidong)
Three Row Paint Brush,Brush Oil Photoshop,Brushes Medibang Paint,Fence Paint Brush
Laizhou Chenke trading Co., Ltd. , https://www.chenkegroup.com