State-Based Model Targets - Beta

Use State-Based Model Targets to track an object through a process where parts are added, removed, or manipulated.

State-Based Model Targets have many applications and greatly simplify the authoring of Augmented Reality (AR) experiences for complex objects and experiences. Use this feature to continuously track an object capable of being in different states and configurations.

Use cases

Use the State-based Model Target feature to track an object throughout its assembly or disassembly, for example, a vehicle with and without an attached accessory or an object with movable parts, such as tracking a car with an open or closed door.

Interested? Reach out to vuforia-feedback@ptc.com to join the beta.

Prerequisites

The following assets are necessary to create State-Based Model Targets.

  • A 3D model file with an embedded process plan (any supported 3D model format) or a Creo Illustrate Procedure exported to PVZ. See Illustration Files and Process Plans for State-Based Model Targets to learn how such files are prepared.
  • The Model Target Generator (MTG) desktop tool (version 10.22 or later).
  • Vuforia Engine SDK (Unity or Native).

Introduction

State-Based Model Targets is a feature to reliably detect and track objects that can be in any of a known set of discrete states, which can differ by the presence of parts or the placement of parts. The full set of possible states must be provided when generating a State-Based Model Target.

States are taken from the embedded information in a CAD model file when a State-Based Model Target is created in the Model Target Generator (MTG). The states can be the steps in a procedure in Creo Illustrate or nodes in a process plan. The Vuforia Engine can then detect and track the object in any of its configurations that are specified by the states.

Creating State-Based Model Targets

State-Based Model Targets can be created in two ways:

Advanced State-Based Model Targets need to be trained in the secure Vuforia cloud. This will enable automatic recognition of the physical object and the state it is in. More precisely, the trained neural network will pick a state for tracking that closely matches what is currently seen.

NOTE: Advanced State-Based Model Targets are optimized to provide reliable tracking no matter which state the object is in. Although Vuforia Engine outputs the currently detected state, we strongly discourage using this information to validate whether the user has performed the step correctly. Advanced State-Based Model Targets do not verify the exact state and are tuned to pick a state best suited for tracking; this might not take into account smaller differences or differences that are not visible from the current position and, therefore, not be picked in consecutive order.

Standard (non-trained) State-Based Model Targets can be generated without training but require the application to set the current state at runtime manually. This can, e.g., be done from a procedure player that allows the user to switch through the steps of an assembly procedure. The corresponding state could be set for the model target to ensure it can be detected and tracked in the current state.

Can this page be better?
Share your feedback via our issue tracker