This chip is less than 1 square centimeter processing 2 billion images per second

 Penn Engineering Today confirms that a team of researchers, including Farshid Ashtiani, Alexander J. Geers and Firooz Aflatouni, have successfully developed this chip with an area of ​​less than 1 square centimeter.

This chip can perform whole image classification in about half a nanosecond without the need for a separate processor or memory.

In traditional AI systems used for image recognition, the image of the target object is first formed on an image sensor, such as a digital camera in a smartphone. The image sensor then converts the light into an electrical signal and finally binary data. Only then will the system be able to 'understand' the image enough to process, analyze, store and classify the images through the computer chip.

This chip has an area of ​​less than 1 square centimeter.

Although today's digital chips can perform billions of calculations per second, more complex image classification such as identifying moving objects or 3D object recognition is pushing ahead. beyond the limits of even the most powerful technologies.

The current rate limits of these technologies are established from the clock-based scheduling of computational steps in a computer processor, where calculations take place on a linear schedule one after another.

Penn State engineers have created the first scalable chip that makes classification and image recognition near-instant, by designing a solution that eliminates the most time-consuming aspects in traditional chip-based AI image processing.

Their custom 9.3mm square processor directly processes light received from the 'subject of interest' using what they call an 'optical deep neural network.'

The researchers' processor effectively makes use of 'optical neurons' connected by optical wires, or waveguides, to form a deep network of multiple layers.

Thanks to the chip's ability to compute as light passes through it to directly read and process optical signals, the researchers claim that the chip does not need to store information and can perform complete classification. set of images in about half a nanosecond.

This chip does not need to store information and can perform classification of an entire image in half a nanosecond.

"We're not the first to come up with technology that reads optical signals directly, but we're the first to create a complete system in a chip that is compatible with existing technology," says Geers. is both extensible to work with more complex data.'

The team hopes this chip will be used in automatically detecting text in images, helping self-driving cars recognize obstacles and other computer vision tasks.

AI has been changing the world of camera technology in recent months. Earlier this year, scientists developed an AI camera that can capture full color in complete darkness.

This week, Camero-Tech announced their latest AI-powered recognition system, called Xaver 1000, which allows soldiers to see through walls before attacking.