Glossary Terms

Model robustness is a model’s ability to maintain its specified level of performance under differing circumstances, such as a different scanner, site, or patient population. Source: FDA Digital Health and AI Glossary Read more

Model weights are the learnable parameters that encode an AI model’s core capabilities, adjusted as the model is trained. Source: FDA Digital Health and AI Glossary Read more

Motion tracking captures and digitizes physical movement in real time, mapping a user’s real-world actions onto corresponding movements in a virtual environment. Source: Unity Glossary Read more

Motion-to-photon latency is the total delay between a user’s physical movement and the moment that movement is reflected on the display. Keeping it low, ideally under 20 milliseconds, is critical to preventing discomfort. Source: Unity Glossary Read more

The multiverse refers to multiple digital worlds interconnected or operating in tandem, extending the idea of parallel connected realities. Source: ACM Metaverse Glossary Read more

A neural network is a computational model inspired by the structure of the human brain, made up of interconnected nodes, or neurons, organized into layers. Source: FDA Digital Health and AI Glossary Read more

Object telepresence is the subjective experience that a virtual object is genuinely situated in the real world. Source: ACM Metaverse Glossary (Borg, Johansen, Thomsen & Kraus, 2012) Read more

OpenXR is an open, royalty-free standard from the Khronos Group that provides a common interface between applications and the various VR and AR platforms. It reduces fragmentation, allowing an application to run across multiple hardware ecosystems without significant rework. Source: Unity Glossary Read more

Outside-in tracking determines a headset’s position using fixed external sensors placed around the user, which requires preparing the surrounding space in advance. Source: XRSI Standard XR Definitions (XR-001) Read more

Overfitting occurs when a model learns its training data too thoroughly, capturing not only the underlying patterns but also noise and random fluctuation. Such a model performs well on familiar data and poorly on new cases. Source: FDA Digital Health and AI Glossary Read more