Optical illusions confuse and confound the eyes, but psychologists still aren’t quite sure how or why. While they understand how the various components of the eye react to these impossible images and have concluded that optical illusions work on most people, some questions remain.

86% of people experience “illusory perception errors” when they stare at this image, the subject of research published in ‘Frontiers in Human Neuroscience.’ 

Recent research has helped reshape our understanding of optical illusions. Experts no longer believe optical illusions are mere signs of imperfections in our perceptive powers. Rather, they may be the result of unsuccessful “experiments.” 

One study, published in Frontiers in Human Neuroscience focused on the ways the pupil responds to illusory sights. It found that 86% of individuals experienced the results of failed “experiments” when they looked at the image above.  

“Illusions exist,” says AI researcher Robert Max Williams, “because our brain and eyes are performing messy and ad-hoc processes to extract a visual scene from an otherwise incomprehensible light field, created by a physical world which we are almost completely sealed off from.” In a sense, the presence of inexplicable optical illusions is as fundamental a part of seeing the world as any other.     

Optical Illusions and Computer Vision

Researchers like Williams believe optical illusions could prove a useful resource for their efforts to develop solutions capable of making the same subtle predictions and adjustments as the human eye. If scientists want to achieve “General Vision,” and truly mimic the human eye with artificial intelligence, teaching machines to recognize optical illusions and mimic a human response may play an essential role. 

“I don’t think it’s possible,” Williams adds, “to make a visual system expressive enough to be considered ‘perception’ which is also free from illusions.” 

Williams contributed to an open source dataset of more than 6,700 optical illusions in 2018. The mission was partially inspired by the revelation that existing computer vision models were incapable of recognizing optical illusions or creating their own

Initial attempts to create these conditions weren’t promising, but Williams, his collaborators, and the AI engineering community at large are still committed to research involving optical illusions. A breakthrough would mean computer vision models capable of going beyond recognizing relevant objects or situations by reasoning. In time, it could help scientists both better understand human visual processes and better recreate them with AI in enterprise settings.  

Why Computer Vision? 

When organizations invest in computer vision and transform their approach to visual data management and analytics, they quickly recognize game-changing benefits. 

  • Identify opportunities for process improvement: Where are bottlenecks, inefficiencies, and waste affecting your production processes? By training computer vision models to recognize inaccuracies or deficiencies in production, teams can automate a more effective response. 
  • Spot trouble proactively: Deploying vision AI and video analytics helps organizations across a range of industries proactively detect potential hazards like VOC leaks, packaging defects, and more.
  • Keep staff and customers safer: By enforcing health and safety standards and monitoring restricted zones, vision AI can go beyond traditional approaches to provide for safer facilities.  
  • Get to know your customers: Retailers, restaurants and other enterprises can apply computer vision models to generate customer insights related to wait times, preferences, people flow, and more to improve service. 

See More

Subscribe on Medium to have Plainsight’s weekly updates sent directly to your inbox. Interested in learning more about how Plainsight’s team can help transform your organization’s approach to computer vision and uncover powerful insights? Schedule a conversation today.

 View All Blogs