The Visual Object Identification Composite AIM is specified in the following six sections.

1     Functions of Visual Object Identification

2     Reference Architecture of Visual Object Identification

3     Input/output data of Visual Object Identification

4     Functions of AI Modules of Visual Object Identification

.5     I/O Data of AI Modules of Visual Object Identification

6     AIM Specification and JSON Metadata  of Visual Object Identification

1      Functions of Visual Object Identification

The purpose of the Visual Object Identification (OSD-VOI) AIM is to produce the Identifier of a Visual Object in a Visual Scene populated by a plurality of Objects that an Entity – the digital representation of a human or an Avatar – indicates by pointing at one of them with a finger.

2      Reference Architecture of Visual Object Identification

Figure 1 depicts the AIM implementing the Visual Object Identification AIM.

Figure 1 – Reference Model of the Visual Object Identification AIM

The workflow of the Visual Spatial Object Identification unfolds as follows:

  1. Direction Identification provides the (ϕ,θ) angles obtained by analysing the finger of the human.
  2. Visual Object Extraction uses the Visual Scene Geometry and the Direction to find the Object intersected by the line identified by (ϕ,θ) passing through the Entity’s straight finger. It is assumed that one and only one Object is found.
  3. Object Instance Identification provides the ID of the Object Instance.

3      Input/output data of Visual Object Identification

Table 1 gives the input/output data of Visual Object Identification.

Table 1 – I/O data of Visual Object Identification

Input data From Comment
Body Descriptors Upstream Visual Scene Description There is a human pointing to an object
Visual Object Upstream Visual Scene Description There are many scene objects
Visual Scene Geometry Upstream Visual Scene Description Provides description of the scene
Output data To Comments
Visual Object Instance Identifier Entity or another AIM Human points to one object only

4      Functions of AI Modules of Visual Object Identification

Table 2 gives functions of the AIMs.

Table 2 – AI Modules of Visual Object Identification

AIM Modules
Visual Direction Identification Produces an Anchored Direction with the following features:

  1. The Point belongs to the forefinger of the Entity.
  2. Azimuth and Elevation of the line enable it to cross the Object intended by the Entity.
Visual Object Extraction Singles out the Visual Object indicated by the Entity.
Visual Instance Identification Produces an Instance ID identifying an element of a set of Visual Objects belonging to a level in a taxonomy.

5      I/O Data of AI Modules of Visual Object Identification

Table 3 gives the list of the AIMs with their functions.

Table 3 – AI Modules of Visual Object Identification

AIM Receives Produces
Visual Direction Identification Visual Scene Geometry
Body Descriptors		 Anchored Direction
Visual Object Extraction Visual Scene Geometry
Anchored Direction		 Target Visual Object
Visual Instance Identification Target Visual Object Instance Identifier

6      Specification of Visual Object Identification AIMs and JSON Metadata

Table 4 – AIM and JSON Metadata

OSD-VOI Visual Object Identification X
OSD-VDI Visual Direction Identification X
OSD-VOE Visual Object Extraction X
OSD-VII Visual Instance Identification X