2022/01/28

MPAI defines its 2022 work program

The 16th MPAI General Assembly (MPAI-16) has approved the 2022 work program. The program contains 14 areas of activity: 5 for approved standards and 9 for projects under development.

Approved standards

In 2021 MPAI has approved 5 standards.

Compression and Understanding of Industrial Data (MPAI-CUI) is a full MPAI standard in the sense that it includes the MPAI-CUI Technical Specification, Reference Software, Conformance Testing and Performance Assessment. Conformance is the status of a technically correct implementation. Performance is the status of an implementation that has a sufficient grade of reliability, robustness, replicability and fairness. In 2022 Version 2 (V2) of MPAI-CUI will be developed that will include more risks in addition to the currently supported seismic and cyber risks.

In 2021 the Context-based Audio Enhancement (MPAI-CAE) and Multimodal Conversation (MPAI-MMC) Technical Specifications were approved. In 2022 the V1 Reference Software, Conformance Testing and Performance Assessment and the V2 Technical Specifications will be developed.

For MPAI-CAE, V2 will extend some of the approved use cases, such as Audio Recording Preservation, with new functionalities such as support of other physical media (e.g., vinyl). The technologies of separation of speech from environment sound and classification of different sound components will be developed and applied to several planned use cases such as Audio On the Go, Human Interaction with a Connected Autonomous Vehicle and Mixed-Reality Collaborative Spaces.

For MPAI-MMC, V2 will enhance V1 with more powerful technologies, such as use of word lattices in Speech Recognition, Dialogue Processing and Translation, and will add new technologies, such as semantic graphs, to drive more realistic speaking avatars.

In 2021 MPAI approved AI Framework (MPAI-AIF), an important infrastructure specification allowing AI Workflows composed of AI Modules to be executed in a standard AI Framework. In 2022 MPAI will develop the MPAI-AIF reference software, an enabler of the reference software of other MPAI standards, and the Conformance Testing. As typical of such standards, implementation needs will drive the request for new MPAI-AIF functionalities.

The 5th MPAI standard approved in 2021 – Governance of the MPAI Ecosystem (MPAI-GME) – lays down the rules governing the ecosystem triggered by MPAI standards including the MPAI Store, a non-profit foundation with the task of receiving, verifying and distributing implementations. In 2022, MPAI will establish the MPAI Store Foundation thus kicking off the actual market deployment of MPAI standard implementations.

Standards under development

Currently, MPAI has 9 projects at different level of maturity for most of which a Call for Technologies is planned.

Server-based Predictive Multiplayer Gaming (SPG) uses AI in support of Online Game Servers that are missing or receiving fake controller data from clients.

Mixed-reality Collaborative Spaces (MCS) is a project supporting scenarios where geographically separated humans represented by avatars collaborate in virtual-reality spaces called Ambients. The project integrates technologies from other MPAI groups, e.g.:

  1. Multimodal conversation.
  2. Sound and speech separation.
  3. Avatar representation and animation.

Connected Autonomous Vehicles (CAV) is a project developing a reference model of the CAV that includes the reference models of its subsystems, the AI Modules and all data formats required for their interconnections. The project integrates technologies from other MPAI groups:

  1. Sound and speech separation.
  2. Human-machine communication.
  3. Avatar representation and animation.
  4. Visual object and scene description.
  5. Visual scene presentation and interaction.

Neural Network Watermarking (NNW) addresses a standard to measure, for a given size of the watermarking payload:

  1. The impact, e.g., the degradation of the user experience caused by the watermark applied to a neural network.
  2. The resistance to attacks, e.g., transfer learning, pruning.
  3. The processing cost of watermarking injection, e.g., time, processing cost.

Avatar Representation and Animation (ARA) is a project responding to the needs of several other projects, such as MPAI-MMC, MPAI-CAV and MPAI-MCS, for standard technologies enabling representation and animation of speaking avatars.

End-to-End Video Coding (EEV) seeks to achieve substantial performance improvements with an AI-based end-to-end approach. This is a long-term exploration project.

AI-Enhanced Video Coding (EVC) is developing the MPAI-EVC Evidence Project. The Evidence Project seeks to demonstrate that, by extending or replacing existing coding tools in MPEG-5 EVC with AI tools, compression performance of at least 25% can be achieved. When that goal will be reached, a Call for Technologies will be issued, and standardisation will start.

Integrative Genomic/Sensor Analysis (GSA) uses AI to compress the results of high-throughput experiments combining genomic/proteomic and other data (e.g., video, motion, location, weather, medical sensors).

Visual Object and Scene Description (OSD) addresses the visual data representation needs from different projects.

MPAI meetings in February 2022

Jan-Feb Feb Feb Feb
Group name 31-04 07-11 14-18 21-25 T (UTC)
Governance of MPAI Ecosystem 31 7 14 21 16
AI Framework 31 7 14 21 17
Mixed-reality Collaborative Spaces 31 7 14 21 18
Multimodal Conversation 8 15 22 14
Neural Network Watermarking 1 8 15 22 15
Context-based Audio enhancement 1 8 15 22 16
Connected Autonomous Vehicles 2 9 16 23 13
AI-based End-to-End Video Coding 22 13
9 14
AI-Enhanced Video Coding 16 14
Avatar Representation and Animation 3 10 17 13:30
Server-based Predictive Multiplayer Gaming 3 10 17 14:30
Communication 3 17 15
Industry and Standards 4 18 16
General Assembly (MPAI-17) 23 15