Computer vision and human vision both “look,” but they don’t work the same way. Humans are good at understanding meaning. If you see a person in a messy room, you instantly know what’s important, even with bad lighting or weird angles. Computers usually don’t understand meaning by default. They see the world as numbers (pixels), and they often need clear conditions or extra rules to decide what something “is.” Humans also handle depth, shadows, and changes in lighting very naturally. A camera image is flat, and small changes (a shadow, glare, low light) can confuse the computer a lot more than a person.
To help the computer “see” or track what we care about, we usually make the situation simpler. One technique is controlling the environment: use bright, even lighting and a clean background so the subject stands out. Another is increasing contrast (like wearing a bright color against a dark wall). In code, you can use methods like motion detection (compare frames to find movement), background subtraction (separate the person from the background), thresholding (turn the image into black/white based on brightness), and color tracking (follow a specific color range). You can also add markers on purpose—like a QR-code-style tag, a colored sticker, or a high-contrast shape—so the computer can track that instead of trying to recognize a whole body or face. The general strategy is: don’t make the computer solve a hard problem if you can redesign the setup to make it an easy one.
Computer vision is strongly connected to tracking and surveillance, and that affects interactive art in a few ways. On the positive side, tracking can create powerful experiences: the artwork can “respond” to your body, movement, or presence in a way that feels alive. But the same tracking can also feel invasive, because people know cameras are used to monitor and identify them in real life. That can create discomfort or distrust, especially if the artwork is not clear about what it records. Artists using computer vision should think about consent and transparency: tell the audience what is being tracked, avoid collecting personal data when it isn’t needed, and design interactions that don’t require identifying someone (tracking motion or silhouettes instead of faces). In interactive art, the best use of vision often focuses on participation and expression, not surveillance.