Table 1 lists the Data Types used by CAV-TEC. Each entry includes a link to the relevant specifications. Note that several Data Types are specified by other MPAI Technical Specifications than CAV-TEC: MPAI-CAE, MPAI-MMC, MPAI-OSD, MPAI-PAF, and MPAI-TFA. Data Types referenced by such Data Types are not listed here but are specified by the relevant Technical Specification.
Table 1 – Data Types referenced by CAV-TEC
2 Conformance testing
The Conformance a Data instance conforms with AIH-HSP V1.0 is expressed by one of the two statements:
- “Data conforms with the relevant (Non-MPAI) standard” – for Data.
- “Data validates against the Data Type Schema” – for Data Object.
The latter statement implies that:
A Data instance Conforms with AIH-HSP V1.0 specified Data Type if:
- Its JSON Object validates against its JSON Schema.
- Any included JSON Object validates against its JSON Schema.
- All Data in the JSON Object:
- Have the specified Data Types.
- Conform with the Qualifiers signaled in their JSON Schemas. For example, if the data cldata to be UNICODE, it should conform with what the Text Qualifier (MPAI-TFA V1.4) defines as UNICODE.
Note that at this stage the AIH-HSP V1.0 does specifies Conformance Testing for Data Types.
3 Performance Assessment
Performance is an umbrella term used to describe a variety of attributes – some specific of the application domain served by a specific Data Type. Therefore, Performance Assessment Specifications provide methods and procedures to measure how well a Data instance represents an original Data entity. Performance of an Implementation includes methods and procedures for all or a subset of the following characteristics:
- Quality– for example, how well a Scene Descriptors instance represent a scene.
- Bias: – for example, how dependent on specific features of the training data is the inference represented by the Data instance.
- Legality– for example, whether the Data instance was produced in a jurisdiction at a time by an AIM that complies with the relevant a regulation, e.g., the European AI Act.
- Ethics – for example, the data instance complies to a target ethical standard.
Note that at this stage the AIH-HSP V1.0 specifies Performance Assessment only of some Data Types.