Seeing Machines (AIM:SEE) published a technical paper demonstrating how robots and other Physical AI systems can achieve real-time 3D human understanding directly on embedded hardware.
The paper details the company's Human Mesh Recovery (HMR) technology, which uses a single camera to reconstruct a 3D representation of a person's location, posture and movement, letting machines respond safely in settings such as factories, hospitals and warehouses.
Seeing Machines said the model runs at up to 180 frames per second on NVIDIA's Jetson Thor hardware, processing information locally rather than relying on a cloud connection, and works with both standard RGB and RGB-D cameras without adding to its parameter count.
The company launched its Physical AI Platform in August, extending its Human-Centred AI approach from vehicles into robotics, following an Advanced Development Contract with a global industrial technology company to build a factory-automation proof-of-concept.
That platform draws on more than 25 years of human factors research underpinning Seeing Machines' existing automotive and aviation businesses, where its Driver and Occupant Monitoring System is now fitted to roughly 6.1 million cars on the road.
HMR is pitched as a core capability within that platform, targeting deployments across manufacturing, logistics, healthcare, aged care and mining.
"Our HMR technology delivers a compelling combination of accuracy, latency and model size on embedded hardware, bringing detailed 3D human understanding into real-world environments", said chief executive Paul McGlone.