In “Computer Vision for Artists and Designers,” Golan Levin explains that computers do not see like humans. I realized that computer vision is not really vision. It is measurement.
When I look at a room, I see people, emotions, and context. A computer sees pixels. It compares brightness values. It detects motion between frames. It subtracts a stored background. Techniques like frame differencing and thresholding reduce the world to contrast and change. The computer does not understand meaning. It only processes differences in light and color.
This difference forces artists to redesign the environment. If the lighting changes, the system fails. If a person stands still, motion detection fails. So artists add backlights. They use white walls. They create strong contrast. In Videoplace by Myron Krueger, participants stand in front of a bright background so their silhouettes are easy to track. The interaction feels magical. But it depends on careful staging for the machine.
I was also struck by works about surveillance. In Sorting Daemon, David Rokeby reduces people to color patches. The system does not know who they are. It only sorts hues and shapes. This makes me uncomfortable. It shows how easily people can be turned into data.
For me, the most important insight is this: computer vision changes not only how machines see us, but how we design spaces for machines to see.