YoloOSC
Live object detection with YOLO, sending results over OSC for interactive installations.
What it does
Runs YOLO object detection on a live camera feed and broadcasts detected objects (class, position, confidence) over OSC. Built with openFrameworks for use in interactive art and museum installations at ZKM.
Key features
- Real-time detection — processes live camera feeds at interactive frame rates
- OSC output — detected objects sent as OSC messages for integration with any OSC-capable software
- Multiple YOLO versions — supports YOLOv3, v4, and v5 model weights
- Region of interest — configurable detection zones to focus on specific areas
- Visual debugging — on-screen overlay showing detected objects and bounding boxes
Tech stack
C++ application built on openFrameworks using the ofxTensorFlow2 addon for inference. Communicates via OSC using ofxOsc.
Museum context
Deployed in ZKM exhibitions for visitor interaction tracking — detecting when visitors approach specific artworks or interact with physical objects.