When I compare computer vision to human vision, I realize how differently they work. For me, vision is immediate and layered with meaning, but for a computer, an image is nothing more than an array of numbers, vectors, and pixels. This reading reminded me of the summer I worked on an image understanding project with our faculty in the cybersecurity domain. I saw firsthand how computers don’t “see” an image the way I do. They only interpret numerical values and computations. Because of this, it’s even possible to generate two different sets of numbers that look the same to a computer, but when drawn out, they are hardly identical. That gap between what humans and computers perceive really fascinated me.
To help computers track what we want, techniques like frame differencing, background subtraction, and brightness thresholding come into play. Through this reading, I learned how important the setup is: lighting, camera placement, and reflective markers can make or break the system’s accuracy. It’s almost like computers need us to simplify reality so they can process it.
Where I feel conflicted is in the area of surveillance. In my cybersecurity project, I came to appreciate how powerful these systems can be, but also how much risk comes with that power. Since computers only “understand” numbers, they can be tricked, but they can also be used to monitor, record, and categorise people in ways that feel invasive. In interactive art, I see these tools as playful, engaging, and thought-provoking. But outside of art, they can become a threat. Surveillance powered by computer vision has the capacity to invade privacy, consolidate control, and even manipulate what we think of as reality. That tension makes me think harder about not just what computers can see, but what we allow them to see.