TensorFlow.
TensorFlow.js + MediaPipe are how Remiam runs ML at the edge — on iPads, in retail kiosks, at events. Pose detection at ~30 fps without needing a backend. The studio uses the WebGL backend for desktop class hardware and WASM where compatibility matters more than peak throughput.
The Discovery Reef
A pose-tracked, fish-eye-camera-captured kids activation Remiam built for a UK shopping-mall family-activation tour. Children played a sequence of physical activities inside the activation zone, the camera and TensorFlow-powered pose detection recorded their session, and parents scanned a QR code to give permission and unlock the recorded clip for viewing online through a hosted family platform.
Trophy Tour
A multi-city UK activation tour built to build hype around a major football tournament. Pose-tracked keepy-uppy, an endless Mexican-wave loop where each visitor adds their wave to a growing crowd, a count-down-and-record goal celebration composited into a real goal scene, a studio-experience zone, and a Firebase-backed viewing platform where visitors could find and share their clip afterwards.