From Simple Shape Detection to Building Artificial Visual Systems

From Simple Shape Detection to Building Artificial Visual Systems

One of the guest speakers who spoke at the Neuroscience Seminar at Loyola University Chicago this fall 2025 was Dr. Nicholas Baker. This professor stood out with his presentation on shape detection. He discussed how humans identify shapes and what makes a shape “good” or “bad,” connecting his ideas to Gestalt principles of perception. In his research study, participants were asked to identify shapes made of dots that were placed against a noisy background. The results showed that curved shapes were recognized quicker and more accurately than angular shapes, suggesting that smooth and continuous forms are easier for our brains to process. Dr. Baker explained that the “best” shape is the circle because it has a consistent radius and maximizes the ratio of volume to perimeter, making it efficient and visually simple for the brain to detect. Triangles and squares are also easy to detect because they have a small number of vertices and still maintain a large volume compared to the perimeter. This finding supports the Gestalt principle of continuity, which describes how humans tend to perceive connected and flowing pieces as part of the same object. Overall, his talk highlighted how our brains organize visual information and why certain shapes feel more familiar or pleasing to us than others.

Before the seminar, we read “The Role of Vertical Mirror Symmetry in Visual Shape Detection” by Machilsen, Pauwels, and Wagemans, which examined how symmetry influences figure–ground segregation. The authors found that people were able to identify shapes with symmetry more often than random ones, showing that vertical symmetry helps the brain find shapes in messy or noisy images. However, Dr. Baker’s research showed that grouping cues like curved lines and smooth connections had a larger effect on shape detection than symmetry. In his own findings, symmetry did not play as strong a role as the volume-to-perimeter ratio, suggesting that shape geometry might guide visual perception more than symmetry alone. 

This topic connects well to recent work at Harvard Medical School, where Dr. Carlos Ponce, an assistant professor of neurobiology, studies how the brain processes shapes to build and improve artificial intelligence (AI) vision systems. Ponce’s research combines computational modeling with electrophysiological recordings from macaque monkeys to understand how neurons encode complex visual information. The goal of his lab is to map the specialized networks of the brain responsible for tasks like recognizing shapes, faces, or tracking movement. By gaining insight into how the brain encodes this data, researchers can build computational models that mimic human vision. Computational systems that can “see” and interpret objects as efficiently as (or even better than) human visual systems, much like Dr. Baker’s findings describe (Harvard Medical School, “Seeing Shapes”). 

Both Dr. Baker and Dr. Ponce’s work highlights how deeply our brains are wired for pattern recognition. From understanding how the human brain detects curved outlines in noisy images to building artificial systems capable of human-like perception, there is so much more to learn and improve upon in neuroscience. Understanding how the brain simplifies the world into shapes explains why certain forms feel intuitive or pleasing but also helps bridge neuroscience with increasingly advancing technologies in AI and computer vision.

References

 Machilsen, B., Pauwels, M. and Wagemans, J. (2009) The role of vertical mirror symmetry in visual shape detection, The role of vertical mirror symmetry in visual shape detection | Jov | Arvo journals. Available at: https://jov.arvojournals.org/article.aspx?articleid=2122115.

 “Seeing Shapes.” Harvard Medical School, 6 May 2019, https://hms.harvard.edu/news/seeing-shapes.

Baker, Nicholas. Neuroscience Seminar, Loyola University Chicago, Fall 2025. Lecture.