VAS innovates by bringing together key technologies across the government cloud service provider onto a single coherent architecture for the implementation of the platform. Various AI/Cognitive, Machine learning services across government cloud providers such as AWS and Azure were used together with the deployment of custom in-house video analytics models both for facilitating model development as well as operational inferencing.

VAS is designed to support secure exchange of video data and metadata allowing greater availability to data on a standardized basis across the organization in a managed and secure way. Based on a common need across agencies to prototype and envision future operational workflows to drive change in operational strategies, VAS also provides user friendly prototyping features as well as security compliant and scalable deployment architecture to allow agencies to develop new custom video analytics capabilities without the need to setup multiple environments.

VAS has been launched for adoption across public agencies as an initial pilot phase at the end of 2020 offering a range of in-house video analytics capabilities. Numerous agencies have been onboarded to use the pilot platform for trials, providing insights to be used for policy development as well as more efficient deployment agency operations.

Based on adoption of the initial implementation, learnings were derived from new agencies use cases and feedback, which continue to be applied in continuous innovation of platform features. The platform is now undergoing a new expansion phase based on agency feedback to enable features addressing wider government use cases in the development and deployment of video analytics models.

Building on industry partnerships, the current phase also introduces new capabilities to meet agencies' needs and expand platform synergies through agencies’ operational integration into VAS, and facilitate wider adoption of video analytics capabilities.

Government officials were continuously involved in the innovation process through identification of use cases, technical solutioning as well as system trials, providing continuous feedback and facilitating development of in-house AI models relevant and aligned to address agency use cases.

Collaborations also involved partnerships with service providers on the government cloud technology stack, such as AWS and Azure, providing close guidance and support on system architecture and implementation.

Several archetypes of government officials were key user beneficiaries:

• Operations managers benefits from on-demand access to a platform to process live or recorded video feeds for operations work.
• Policy planners, can reference aggregated data extracted from video such as for pedestrian studies on public walkways
• Technically inclined users have the ability train custom models.
• Local companies offering AI/ML technology benefits from leveraging the platform for reach across agencies.

VAS brings greater cost effectiveness across whole-of-government by allowing multiple agencies to leverage a central platform, avoiding the cost of setting up separate systems, and reaping the benefits of scalability.

The provisioning of VAS allows public agencies to access the same capabilities for less than 50% the outlay compared to if they were to acquire the capabilities individually. With the availability of VAS, multiple agencies have been onboarded for having integrated or performed trials with VAS, translating into cost savings for each agency onboarded.

Agencies on-boarded also see a significant savings in time-to-deploy as they do not need to procure, develop a separate setup from scratch. Results have been measured common product scorecard through metrics indicative of adoption and transactional trends. In future, we expect to deepen partnership with the industry to drive the integration and development of more capabilities to enable platform synergies and drive adoption

A challenge lies in handling disparate video protocols for video management system and data repositories across public sector landscape due to diversity in procurement landscape and differing environments and needs across agencies and their internal teams.
To overcome the challenge, VAS took the approach to harmonize the exchange of video and metadata by providing common standards across public agencies and allowing data to be more consistently accessed for analytics while maintaining government standards for data access.

There is also a need to ensure conformance to government cloud architecture and security requirements as standards are evolving and require adaptation should new architecture or policies be introduced. This is also addressed through the design of VAS as a platform providing a stable foundation, maintained to ensure conformance to government architectural requirements while allowing standardized integrations with other agency systems

• Supporting infrastructure and services in the form of a common technology stack and services that serves the whole of government is important in enabling the development of the platform in scalable and secure manner.
• Policies and rules governing the implementation of video systems as well as data sharing across the organization need to work in tandem with implementation of common video metadata standards in order to facilitate data transfer at scale.
• Leadership and guidance is needed to proliferate the innovation, hence a core capability centre should be empowered to provide consultancy and guidance to agencies in adoption of platform and metadata standards.
• Human and financial resource needs to be assigned in order to perform necessary development of the platform and enable the innovation it facilitates.
• Personal values and motivation of agencies users will need be aligned through platform adoption towards digital transformation and evolution of operations at organizational scale

Yes, similar types of platforms have been developed for structured data sets in private industry as well as public sector to leverage organisational efficiency in the utilization of AI . The potential in this case lies not in replication of the platform itself but allowance of opportunity for agencies to develop their own innovation in training of custom AI models relevant to their problem sets.

One lesson is the need to drive understanding of the ecosystem, beyond targeted agencies, as approach to drive the feature development process. Particularly with AI innovations, as user maturity evolves with the environment, the team requires agility for integration of new features relevant to new needs arising from agencies. As such, another lesson is the need to retain expertise and control over the core architecture and implementation of core features, while allowing flexibility for features to be integrated from other in-house or partner solutions.