what is observation tracking?

Observation tracking, as discussed in the context of WWDC sessions, primarily refers to the tracking of subjects within a scene using advanced algorithms and machine learning. This involves identifying and maintaining focus on the most relevant subjects, such as faces or objects, based on various attributes like movement, speech, and proximity to the camera.

For example, in the session "What’s new in DockKit", intelligent subject tracking is described as a method to analyze a scene in real-time, identifying main subjects and determining the most relevant person to track. This is achieved through a new intelligent tracking pipeline in iOS 18, which processes data from a multiperson tracker and uses a machine learning model to analyze attributes like body pose, face pose, attention, and speaking confidence.

Additionally, in the session "Create enhanced spatial computing experiences with ARKit", object tracking is discussed in the context of ARKit, where it involves getting the position and orientation of items to anchor virtual content to them. This requires supplying a set of reference objects, which encode spatial features of known real-world items.

Here are the relevant sessions mentioned:

1. What’s new in DockKit
2. Create enhanced spatial computing experiences with ARKit
3. What’s new in Create ML
4. Introducing enterprise APIs for visionOS