AIAM ResearchAutomotive Additive Manufacturing

3D printed AI device recognizes objects at the speed of light

UCLA's low-cost 3D printed artificial neural network could give autonomous cars faster response times

Engineers from the University of California, Los Angeles (UCLA) have successfully developed an artificial neural network using 3D printing. The physical device is reportedly capable of analyzing large amounts of data and identifying objects at the speed of light and could have important applications in autonomous vehicles and more.

Developed by a team of electrical and computer engineers, the small 3D printed device may not look like much on its own. In fact, it looks quite simply like a series of small square polymer sheets lined up in front of one another. But these polymer layers (measuring 8 x 8 cm) are really something special.

Let there be light

As the researchers explain in a press release, while many devices today have the ability to identify objects, they mostly rely on cameras or optical sensors to capture the initial data of the object. The 3D printed AI network, however, utilizes a “diffractive deep neural network” which is capable of identifying an object simply by the light bouncing off of it. This means that it can identify the object as quickly as it would take an optical sensor-based device to simply “see” it.’

Moreover, because it does rely on optical sensors, the UCLA device does not require any advanced computing programs to process images of the object. Since the artificial neural network uses diffraction of light to understand its surroundings, it also does not consume any energy.

“This work opens up fundamentally new opportunities to use an artificial intelligence-based passive device to instantaneously analyze data, images and classify objects,” explained Aydogan Ozcan, principal investigator of the study and Chancellor of Professor of Electrical and Computer Engineering at UCLA. “This optical artificial neural network device is intuitively modelled on how the brain processes information. It could be scaled up to enable new camera designs and unique optical components that work passively in medical technologies, robotics, security or any application where image and video data are essential.”

Know when to “stop”


In terms of applications, the UCLA researchers say the AI device could be adapted to speed up data-intensive tasks which necessitate identifying and classifying objects. One important example of this is in autonomous vehicles, which much be able to rapidly “see” and recognize their surroundings. The new artificial neural network could allow for faster response times than are currently possible. If a driverless car encountered a stop sign, for instance, it could see it as soon as light from the sign would hit it.

In the medical field, the technology could also be adapted for use in microscopic imaging. More specifically, it could help to identify signs of disease present in millions of cells.

3D printing AI

In making the artificial neural network, the researchers first came up with a design for a series of thin, 8 cm2 polymer wafers. The wafers, designed to be 3D printed, integrate an uneven surface which diffracts light in various directions. Interestingly, the researchers say that submillimeter-wavelength terahertz frequencies of light can travel through the 3D printed layers (each composed of tens of thousands of artificial neurons) which is why there are multiple layers.

The multiple layers then are what compose the optical network that shapes “how incoming light from the object travels through them.” The identification of the object occurs when the light coming off the object is diffracted to a single artificial neuron (or pixel) which is assigned to that type of object.

The next step was to then “train” the neural network using a computer to identify certain objects by learning the specific pattern of diffracted light. This complex stage relied on deep learning, a type of AI which consists of training machines through repetition and pattern recognition.


“This is intuitively like a very complex maze of glass and mirrors,” Ozcan explained. “The light enters a diffractive network and bounces around the maze until it exits. The system determines what the object is by where most of the light ends up exiting.”

At this stage in the development, the 3D printed neural network has successfully identified handwritten numbers and clothing items. These tests consisted of placing an image in from of a terahertz light source and letting the device identify the images. The device has also been shown to be capable of acting as a lens that projects the image of an object to the other side of the optical network it is in front of. This use, says the research team, is similar to a camera lens but relies on AI rather than physics.

A neural network for under $50

One of the most impressive and unlikely aspects of the innovative AI device is its cost. Reportedly, the device created by the UCLA team cost less than $50 to produce. Further, because it consists of 3D printed components, the artificial neural network can be scaled up to include bigger and/or more layers, increasing the capacity of the network significantly. By adding more artificial neurons to the device, the researchers say it could identify more objects at once or analyze even more complex data.

Moving ahead, the research team says its device could also be adapted for use with visible, infrared or other light frequencies. The 3D printed wafers, for their part, could also feasibly be swapped for layers made using lithography or other printing methods.

The potentially disruptive AI device was recently published about in the journal Science.


Tess Boissonneault

Tess Boissonneault moved from her home of Montreal, Canada to the Netherlands in 2014 to pursue a master’s degree in Media Studies at the University of Amsterdam. It was during her time in Amsterdam that she became acquainted with 3D printing technology and began writing for a local additive manufacturing news platform. Now based in France, Tess has over two and a half years experience writing, editing and publishing additive manufacturing content with a particular interest in women working within the industry. She is an avid follower of the ever-evolving AM industry.

Related Articles

We use cookies to give you the best online experience. By agreeing you accept the use of cookies in accordance with our cookie policy.

Privacy Settings saved!
Privacy Settings

When you visit any web site, it may store or retrieve information on your browser, mostly in the form of cookies. Control your personal Cookie Services here.

These cookies are necessary for the website to function and cannot be switched off in our systems.

In order to use this website we use the following technically required cookies
  • wordpress_test_cookie
  • wordpress_logged_in_
  • wordpress_sec

Decline all Services
Accept all Services



Join industry leaders and receive the latest insights on what really matters in AM!

This information will never be shared with 3rd parties

I’ve read and accept the privacy policy.*