One large task of the brain as it relates to interfacing with the surrounding world is creating an accurate image of the surrounding world. This has many diverse yet interconnected subfacets of study, for instance shading, tone, texture, and depth. Mikayla Cutler and Luke Baumel along with other researchers at Loyola University Chicago have compared the brain to computational imaging software, looking for differences as to how 3-Dimensional aspects may influence image recognition in the paper “Beyond the Contour: How 3D Cues Enhance Object Recognition in Humans and Neural Networks” (2025). Stereopsis describes how two eyes may form an image together, taking into account objects such as depth as a key factor. This is the kind of vision they focused on. Vision with utilization of one eye for close objects and another eye for far objects is known as monovision. Monovision as it relates to depth perception was the research topic of a study done by various researchers out of the Center for Vision Research at York University in the paper “Monovision: Consequences for depth perception from large disparities” (2019).
Mikayla Cutler et. al. used what is known as a neural network to simulate the brain computationally (2025). This has many individual nodes that form a complex interconnected web, and communicate in a fashion not dissimilar to how axons from one neuron may innervate many other neurons to form the brain. They input different kinds of images into this neural network, with differences in object texture, shadow placement, and other aspects related to 3-Dimensional image recognition. They also gave these images to people, and compared results. Findings indicate that both the human brain and neural networks strongly rely on shape when it comes to discerning what an image is a representation of (Cutler et. al. 2025). Humans do better when an object is viewed from a traditional perspective, whereas neural networks perform equally regardless of viewpoint. They believe that humans rely on shadows for additional information, whereas a computer would rely on similarity to previously encountered images integrated into its network of nodes (Cutler et. al. 2025).
Another aspect used in interpreting the world is depth. It provides cues as to how far an object may be from a particular location to an individual. Smith et. al. use the previously established basis that stereopsis was lower in monovisual individuals, though since not much else was known, they looked into the various factors that might play a role in depth perception differences (2019). They tested individuals in two age groups with 20/20 eyesight who had typical utilization of both eyes on two occasions, one time adding a lens to an eye to make them monovisual. They tested for how well objects could be made out at various distances, and found that both young and old age groups scored lower for farther distances. After performing dynamic photorefraction, it was found that the younger group would sacrifice resources from one eye to aid the other, while the older group would maintain a neutral balanced level of performance in both eyes (Smith et. al. 2019). This varies from non-monovisual individuals, who could focus the eyes on objects at all tested depths and recognize the objects regardless of distance.
Much information can be derived on how the standard human brain works to visually interpret the world around it by comparing it to various nonstandard models. Because of comparison to neural networks, image recognition can be seen to be shape based and improved from traditional angles (Cutlet et. al. 2025). Looking into depth perception of monovision demonstrates how the default brain adjusts to surroundings to estimate how far an object may be from a given location (Smith et. al. 2019). The more research is done into divergences from the normal brain, the more can be understood about the natural human brain condition.
References
Cutler, M., & Baumel, L. D., Tocco, J., Friebel, W., Thiruvathukal, G. K., & Baker, N. (2025). Beyond the contour: How 3D cues enhance object recognition in humans and neural networks. Journal of Vision. (Pre-publication draft)
Smith, C. E., Allison, R. S., Wilkinson, F., & Wilcox, L. M. (2019). Monovision: Consequences for depth perception from large disparities. Experimental Eye Research, 183, 62–67. https://doi.org/10.1016/j.exer.2018.09.005
No comments:
Post a Comment