What is Visual Intelligence?
The Visual Intelligence Technology Stack
The problems that have to be addressed to enable Visual Intelligence applications are complex. Elastic computing and storage capacity are prerequisites. Infrastructure as a Services (IaaS) is provided today as a commodity by Microsoft Azure, AWS, GCP and others. In addition to the cloud, multiple services are necessary to develop, deploy, and operate AI and IoT applications. Hence the need for a specific and complete technology stack that overcomes the following challenges:
Data integration: The data inflow should be automated. The platform should tap into every possible information system, or database, available at the company level and create a unified, federated image of all the data they contain. This requires a versatile interface, to handle heterogeneous types of data.
Data indexing: All data needs to be indexed in a way that becomes easily browsable and pollable via context-specific requests. For instance, the database should be capable of answering questions of the type: What was the state of an asset in my infrastructure one year ago?
Storage: Visual data is heavy by essence. It needs to be stored persistently and cost-effectively while maintaining a high level of accessibility.
Data security: Industrial infrastructure data is critical. The architecture of the stack should be secure by design and meet with the highest requirements of enterprise software.
Processing: Automated analyze of visual data requires heavy-duty and specialized processing capabilities. This includes machine learning services to efficiently train and run AI-models.
Visualization: In an industrial context, 3D and timelapse are at the heart of situation awareness. Data should be explored as intuitively as we explore reality. A set of 2D/3D viewers are the center of the UX/UI in order to visualize an enriched version of the digitalized reality. Technologies are compatible with AR/VR to create advanced on-site intervention tools.
Editing tools: Data inspection, dataset annotation for machine learning, reporting, publishing findings and insight about operations, all these tasks require advanced editing tools embedded in the visualization framework. They should be intuitive and accessible to all users to enforce enterprise-wide collaboration.
Developers tools: An organization’s IT development and data science team should be capable of creating and deploying at scale its own applications and data processing algorithms. AI, machine learning and data science competencies should be internalized as they are part of the post-digital transformation core business.
Open, Scalable, Future-Proof: A published and documented SDK and APIs are essential to interface the platform with the whole IT ecosystem of an organization and incorporate any new open source or proprietary software innovations. The techniques used in the most cutting-edge stack today will be obsolete in 5 to 10 years. An AI and IoT platform architecture must provide the capability to replace any components with their next-generation improvements