Beyond the Screen
Computer vision is usually discussed in the context of photo tagging, face filters and AR overlays on phone screens. But the same underlying techniques — detecting, classifying and tracking objects from camera input — apply directly to physical installations that need to understand what's happening in a real space in real time, independent of any display.
Used this way, computer vision becomes a sensing input for mechanical and electronic systems rather than a content layer for a screen. A camera can tell a pneumatic target when to pop up, tell a turret where to aim, or tell a lighting rig how many people are in a zone — all without a participant ever looking at a screen.
Physical-World Applications
Our engineering work on physical attractions has applied computer vision as a direct sensing layer for mechanical and safety systems, not just for on-screen recognition. Related capability underpins our AR development work.
- Object and person detection feeding directly into targeting or triggering logic
- Motion tracking for safety interlocks stopping mechanical movement when someone enters a hazard zone
- Presence and dwell counting driving installation behaviour without wearables or badges
- Real-time trajectory prediction for systems that need to respond ahead of movement, not after it
- Environmental classification distinguishing static background from moving participants reliably outdoors
The Engineering Challenge Differs From Screen-Based CV
Screen-based computer vision usually has forgiving conditions — controlled lighting, a phone held steady, a user cooperating with the system. Physical-world sensing for mechanical systems has none of these luxuries: variable outdoor lighting, moving cameras on moving structures, participants who aren't trying to be recognised, and — critically — real consequences when detection is wrong.
Need real-world computer vision sensing? We offer free consultations on tracking systems... Let's discuss your installation →
Latency and Reliability Matter More
When computer vision drives a mechanical response — a target popping up, a turret adjusting aim, a safety stop triggering — the acceptable latency and false-negative rate are far stricter than for a phone filter that can afford to occasionally miss a frame. This changes camera selection, processing hardware and algorithm choice substantially from typical screen-based CV projects.
Where This Fits in Installation Design
When scoping an installation, it's worth asking early whether computer vision is being used to render something on a screen or to drive a physical response, because the two problems demand different hardware, different testing regimes and different safety considerations, all of which we account for through our partnership development process.
