X
Innovation

Virtual robots fooled by visual illusions

Researchers at University College London (UCL) have written a computer program using neural networks which are duped by optical illusions the same way as we do. Their virtual robots, which were trained to 'see' like us, could help to understand why we fall for optical illusions. This might also be important for robot vision. If robots are trained to 'see' like us, they will act like us -- and make mistakes. Very interesting...
Written by Roland Piquepaille, Inactive

Researchers at University College London (UCL) have written a computer program using neural networks which are duped by optical illusions the same way as we do. Their virtual robots, which were trained to 'see' like us, could help to understand why we fall for optical illusions. This might also be important for robot vision. If robots are trained to 'see' like us, they will act like us -- and make mistakes. Very interesting...

A visual illusion from Beau Lotto

You can see above one of the visual illusions featured on the Web site of the lead researcher for this project. Here is the text above the illusion. "The central squares of the two discs below (see black dots) appear very different in colour: Green on the left and orange on the right. Despite this appearance, the surfaces are in fact physically identical. Move your mouse over the 'mask' to reveal their 'true' similarity." (Credit: Beau Lotto) Of course, moving your cursor on the 'mask' on the image above will not work. You need to visit the real illusion here (the 'mask' is on the top right).

This research work has been driven by Dr. R. Beau Lotto, a Lecturer in Neuroscience at the UCL institute of Ophthalmology . He worked on this project with David Corney. Please visit his lottolab site for more details.

Now, let's take a look at the UCL news release to discover what is a visual illusion. "Illusions are defined in the Oxford English Dictionary as 'something that deceives or deludes by producing a false impression.' Visual illusions, such as the 'Hermann Grid Illusion,' trick the viewer into misinterpreting –- in this case –- shades of grey. Dr. Beau Lotto said: 'Sometimes the best way to understand how the visual brain works is to understand why sometimes it does not. Thus lightness illusions have been the focus of scientists, philosophers and artists interested in how the mind works for centuries. And yet why we see them is still unclear.'"

So what did the researchers do? "To address the question of why humans see illusions, researchers at the UCL Institute of Ophthalmology used artificial neural networks, effectively virtual toy robots with miniature virtual brains, to model, not human vision as such, but human visual ecology. Dr David Corney in Dr. Lotto's lab trained the virtual robots to predict the reflectance (shades of grey) of surfaces in different 3D scenes not unlike those found in nature. Although the robots could interpret most of the scenes effectively, and differentiate between surfaces correctly, they also -- as a consequence -- exhibited the same lightness illusions that humans see."

In "Artificial brain falls for optical illusions," New Scientist provides additional details. "The brain learns how to tackle this through trial and error when we are babies, the theory goes. Mostly it gets it right, but occasionally a scene contradicts our previous experiences. The brain gets it wrong and we perceive an object lighter or darker than it really is -- creating an illusion. Until now there has been no way of knowing whether this theory is correct. Beau Lotto and David Corney at University College London, UK, think they have finally done it."

And how did they work? "They created a program that learns to predict the lightness of an image based on its past experiences -- just like a baby. And just like a human, it falls prey to optical illusions. They trained it using 10,000 greyscale images of fallen leaves that animals might face in nature. It had to predict the true shade of the centre pixel of the images, and change its technique depending on whether its answer was right or wrong. The researchers then tested the program on lightness illusions that would fool humans. First, it was shown images of a light object on a darker background, and vice versa. Just like humans, the software predicted the objects to be respectively lighter and darker than they really were. It also exhibited more subtle similarities -- overestimating lighter shades more than darker shades.

For even more information, this research work has been published in PLoS Computational Biology on September 28, 2007. Here is a link to this paper, "What Are Lightness Illusions and Why Do We See Them?"

Sources: University College London (UCL) news release, September 28, 2007; David Robson, New Scientist, September 28, 2007; and various websites

You'll find related stories by following the links below.

Editorial standards