summary: Researchers have uncovered key aspects of infant vision, revealing that young infants experience a unique visual diet consisting of simple, high-contrast patterns and edges found in their everyday environments. This “diet” has a great influence on the developmental trajectory.
Their study used head-mounted cameras on infants to directly observe and analyze visual stimuli in their everyday environment, and compared them to adults’ perceptions. These discoveries not only advance our understanding of human visual development, but also provide insights for improving his AI visual system through a similar gradual learning process.
Important facts:
- Unique visual input: Infants are naturally attracted to and surrounded by high-contrast patterns in their everyday environments, which is important for visual development.
- Impact on AI learning: The study’s methodology and findings have been applied to enhance visual systems in artificial intelligence, and the results show that AI trained on a set of images that mimics the visual experience of young children has superior performance. .
- Broader meaning: This study provides a deeper understanding of how early visual experiences are optimized for developmental progress and may lead to better early intervention strategies for visual abnormalities.
sauce: indiana university
What are young children looking at? what are they seeing?
The answers to these questions are very different for the youngest and older infants, children, and adults. These early scenes, featuring a few high-contrast edges in simple patterns, also contain the very material needed to build the strong foundations of human vision.
That’s the result of a new study called “Edge Simplicity Bias in Visual Input to Infants,” published May 10 in the journal scientific progress By IU researchers Erin Anderson, Rowan Candy, Jason Gold, and Linda Smith.
“The first assumption for anyone thinking about the role of experience in visual development is always that, on the scale of everyday experience, visual input is roughly the same for everyone,” said lead researcher and co-author Professor Linda Smith from the department explains. Psychology and brain science.
“However, this study states that visual input changes with development. It is not the same for everyone. For young infants, the input available in daily life appears to be specific to their age. ”
Previous studies in the lab and clinic have shown that young children prefer to see simple, high-contrast scenes with large black stripes or checkers. The current study is the first to ask the extent to which these preferences constitute input into their daily lives.
“To see what young babies see and see,” says Anderson, a former postdoctoral fellow in Smith’s Cognitive Development Laboratory. She and her colleagues fitted infants with head cameras that they wore around the house during daily living activities.
“You can purchase ‘baby flash cards’ for newborns that display these simple, high-contrast images,” she explains.
“What the headcam videos show, and what this work shows, is that young toddlers find these kinds of images all around them in their daily lives, just by looking at lights, corners of the ceiling, etc. That means there is.”
“What we discovered is a very special early ‘diet’ for visual development,” Smith adds. “Just like with food, young children don’t start out with fancy, complicated meals or pizza; they start with simple, developmentally appropriate nutrition.”
Previous research has recognized the critical nature of this early period for the future development of human vision. For example, infants born with visual defects such as cataracts, or infants in orphanages with limited visual experience, have been shown to have lifelong visual impairments.
The current study provides preliminary data to address these deficiencies. This also has important implications for building AI vision systems, which similarly can acquire stronger visual skills by starting with the same simple, high-contrast visual content.
“A huge amount of input into daily life”
To determine the characteristics of visual input in infants aged approximately 3 to 13 months, the researchers placed 10 infants and 10 adult caregivers with head-mounted video cameras and conducted tests at home. Visual recordings of 70 hours of daily life were collected and analyzed.
Clear differences emerge between infant image content and adult images, with infant visual fields having a greater concentration of simpler patterns and higher contrast edges than adult images.
Smith speculates that the reason for this view is not only that infants turn their heads to see features of the world they can see, but also that parents and caregivers are more likely to place objects where they want the infant to see them. doing.
“You have to think about why they’re there. Maybe there’s some unspoken natural knowledge that parents leave toddlers alone where they want to look. If you don’t make a fuss, mom won’t bother you,” she observes.
But is this small group of participants from Bloomington, Indiana representative of a broader group of young children around the world? To answer this question, Smith’s lab conducted the same experiment with collaborators in a small, crowded fishing village in Chennai, India. There, electricity is minimal and much of daily life takes place outdoors.
And while images from head cameras at 6 months and 12 months looked very different from the Bloomington infants, the youngest infants showed high-contrast images in both Chennai and Bloomington. They share a common “diet” of edges and simple patterns.
The bigger picture, past and future
Smith and his collaborators also showed that the same set of images improves the training of an AI vision system.A follow-up to current research, to be published in 2023 Neural Information Processing Systems Conference MinutesThey found that training an AI system by first presenting images characteristic of early infancy improves visual acuity compared to presenting images in a random developmental order or simply presenting images typical of adult daily life. We found that learning to identify images was successful. . A more precise development order produced the best results.
Their research opens up new avenues of evolutionary thought. Smith explains: “One of the things I always asked when I was a graduate student, and perhaps we have the opportunity to answer it, is why motor development in human babies is so slow.
“They spend about three months just listening and watching, and then another six months practicing a little bit of posture and head control. Why are they so slow? The horses come out and race. Yes.”
The study suggests that “over evolutionary time, these slow, gradual, optimized biases work to build highly intelligent visual and auditory systems.” she says. “It’s a story that could be told.”
Meanwhile, their work raises new questions about visual content in early infancy and its role in the development of visual systems, whether human or AI.
Other researchers include Professor Rowan Candy of IU Bloomington’s Department of Optometry and Professor Jason Gold of the Department of Psychological and Brain Sciences.
About this vision and neurodevelopmental research news
author: Liz Rothditcher
sauce: indiana university
contact: Liz Rozdeicher – Indiana University
image: Image credited to Neuroscience News
Original research: Open access.
“Edge simplicity bias in visual input to young childrenWritten by Erin Anderson et al. scientific progress
abstract
Edge simplicity bias in visual input to young children
The development of sparse edge coding in mammalian visual cortex depends on early visual experience. In humans, there are multiple indicators that the statistics of early visual experience have unique properties that may support their development.
However, there is no way to directly measure the edge statistics of infants’ daily life experiences.
Using a head-mounted camera to capture egocentric images of infants and adults in the home, we found that infants’ images had distinct edge statistics compared to adults. For infants, scenes with many sparse edge patterns (few edges or directions) predominate.
The findings suggest biases in early input at the scale of everyday life, perhaps unique to the first few months of life, and provide insight into the quality, quantity, and timing of visual experience during the basic development of human vision. It is something.
