Our eyes are taking in a lot of information when we see something within milliseconds of turning our head to the object in question. How does that happen? What are the internal mechanisms that lead to our being able to differentiate between two shades of green or know the difference between a pen and a pencil?
The unscientific explanation is that there is lot going on inside our brain that makes vision what it is for humans. Of course, it takes science to explain what exactly is going on inside our brains.
Take babies for instance. Within hours of being born, babies fix their gaze on faces. Psychologists at Emory University studied the brains of 30 babies between the ages of 6 to 57 days to learn about the workings of the visual cortex. This was done using functional magnetic resonance imaging (FMRI).
Researchers found that the brains of babies to be adult-like in the framework of the visual cortex is there, as well as patterns of brain activity. The patterns aren’t as strong in babies as they are in adults. The results showed that two regions of the visual cortex associated with the processing of faces fired in sync in the infants, as did the areas associated with the processing of places. In fact, the infant patterns show that both areas are communicating with each other within days of birth.
Understanding how a baby’s brain is organized can help when things aren’t right. For example, if the infant’s visual cortex isn’t well-connected, that could be a biomarker for disorders linked with an aversion to eye contact. “By diagnosing the problem earlier, we could intervene earlier and take advantage of the incredible malleability of the infant brain,” said Daniel Dilks, associate professor of psychology at Emory University and lead author of the study.
The adult visual system is worth studying, as well. Especially since vision is the part of the brain that science understands the best. It was theorized in the early 1900s that the human brain creates images by reversing the process of image formation. Working backwards allows for the brain to infer what kind of face or object would produce that image.
Still, how does the brain do that so quickly? Researchers at the Massachusetts Institute of Technology built a neural network model to show quickly the neural network can infer the features of scene, like a specific face. While most standard deep neural networks used in computer vision label data by indicating the class of an object in the image, the researchers trained the model to reverse the steps by beginning with a two-dimensional image, then adding features such as texture, lighting and so forth. It then creates a 2.5-dimensional image. The 2.5 image specifies the shape and color of the face from a certain viewpoint. The image is then converted to 3D images that doesn’t depend on a particular viewpoint.
The researchers also compared the model’s performance when came to recognizing faces from different viewpoints (i.e. texture was altered or the shape was distorted). The model’s performance was similar to that of a human.
“If we can show evidence that these models might correspond to how the brain works. This work could lead computer vision researchers to take more seriously and invest more engineering resources in this inverse graphics approach to perception,” said Josh Tenenbaum professor of computational cognitive science and a member of MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and the Center for Brains, Minds, and Machines (CBMM) and one of the researchers involved in this study. “The brain is still the gold standard for any kind of machine that sees the world richly and quickly.”