The Silicon Review ^Asia

A team of scientists from the Stanford University have developed a camera that works on AI (Artificial Intelligence).  The camera is designed to navigate robotic cars during any kind of obstacle.

The technology used is based on image recognition that underlies cars and aerial drones depend on AI:  the self-trained computers identify objects like a dog, a person crossing the street or a parked car.

The only problem with the system is that its AI algorithms are too large and slow for using next-generation applications. The scientists’ integrated two different computers to develop an image processor. There are two layers within the prototype camera, i.e. optical computer and digital electronic computer. The optical computer runs with the help of physically processed image data.

After that, the data is filtered in various ways that an electronic computer would otherwise have to do mathematically. The filtering happens through light optics, the layer without any input power. The process saves a good amount of energy and time for the hybrid system.

According to the scientists, the system can be miniaturized for use in aerial drones and hand-held video cameras. After a lot of experimentation, the system was able to recognize aircraft, cars, and pets like cats and dogs within its natural image setting (In both simulations and real-world experiments).

It will be interesting to see if the new invention could really help the autonomous vehicles or not.

“Millions of calculations are circumvented and it all happens at the speed of light. Some future version of our system would be especially useful in rapid decision-making applications, like autonomous vehicles,” said Gordon Wetzstein, an assistant professor at Stanford.