What are some of the ways that computer vision differs from human vision?
The main thing I took from the reading is that computers don’t actually understand what they’re looking at. When I see a room, I just know there’s a person standing there or that it’s night outside. A computer only gets a bunch of pixel values, and it has no idea what any of them mean unless someone programs it to look for something specific. Our eyes also work pretty much anywhere, even in bad lighting or a messy room. Computer vision doesn’t. Levin says no algorithm works in every situation, and if the lighting changes or the background gets busy, it can just stop working.
What are some techniques we can use to help the computer see / track what we’re interested in?
There are a few simple ones. Frame differencing looks at what changed between one frame and the next, which shows movement. Background subtraction compares the camera image to a picture of the empty space, so anything new pops out. Brightness thresholding splits the image into light and dark parts, and you can also just track the brightest spot, like a flashlight. What surprised me is how much the physical setup matters. Using a white background, backlighting, infrared light or reflective materials can make the computer’s job way easier. Sometimes it makes more sense to change the room than to write more complicated code.
How do you think computer vision’s capacity for tracking and surveillance affects its use in interactive art?
I think it works both ways. Tracking is what makes a lot of interactive art fun, because the piece reacts to you and your body, like in Videoplace where your silhouette plays with little creatures. But it’s the same technology used to watch people, so there’s always a bit of an uneasy feeling behind it. Some artists lean into that. In Standards and Double Standards, belts turn to follow you around the room, and Sorting Daemon picks people off the street and sorts them by color. Those pieces made me realize how often we’re tracked without even thinking about it. So for me, computer vision makes interactive art more exciting, but it also means artists have to think about privacy and whether people actually agreed to be watched.