Part 2: How AI Computer Vision is Accelerating and Transforming Asset Inspection 

by | Jun 5, 2020 | Blog | 0 comments

This is Part Two of a two-part series covering how AI can drastically accelerate Digital Transformation in Asset Inspections. Part One can be found here.

As we discussed in the Part One of this series, businesses are currently undergoing a massive digital transformation, creating opportunities to leverage emerging technologies and enable fresh approaches to business practices. One of the areas most impacted is the way businesses are leveraging technologies to accelerate how they perform asset inspection. This transformation requires a strategic approach if they are to deliver bottom line benefits.

Enterprise Asset Management (EAM) is an area of business that is particularly influenced by these advances in technology– with more asset-centric, real-time data available to be fed back into centralized systems. But in order to realize maximum asset availability and benefit from greater business intelligence and a 360-degree, enterprise-wide view, businesses will need the right EAM approach and workflow. One that is geared towards streamlining asset management and maintenance.

One of the key technological driving forces behind the disruptive trends we are seeing today in the digital transformation of asset inspection is the advances in AI based machine learning and computer vision to automate the inspection process at scale.


Enterprises are looking for a way to integrate these steps into a seamless AI workflow such that, when properly orchestrated and operationalized, creates tremendous value. 

In this second part of our series we will cover steps for how Enterprises can operationalize AI and start to realize the value of AI across their operations.  The key steps for operationalizing AI include:

Step 1: Visual data management.  Capture and manage visual data that can be shared and acted on across the organization.

Step 2: AI data labeling.  Comprehensive tools to allow data scientists and subject matter experts to collaborate on data labeling and eliminate data silos.

Step 3: AI Training and Creating AI models.  Seamlessly create AI classifiers and utilize 3rd party AI engines to create AI models for any application.

Step 4: Deploy AI models.  Make the AI models available for business user consumption through an open “AI Marketplace”.

Step 5: AI Visualization.  Visualize AI results through dashboards to quickly search, filter, and integrate into ticketing and other operational workflows.

However, most Enterprises are still relying on disparate point tools to perform these steps. Enterprises are looking for a way to integrate these steps into a seamless AI workflow such that.  when properly orchestrated creates tremendous value. 

Let us explore these steps in more detail to show how Enterprises can start to realize the value of AI in their operations.



Every company is unique and there is no single path for driving AI digital transformation. What is important is that the asset data, AI models, and data visualization analysis are integrated into your operations and workflow acting on a single source of truth.   This means having a comprehensive visual data management platform for creating your digital inspection twin database  and Data Visualization technology allows you to reason over data without having to move your data regardless of where it resides or who needs access.  This eliminates one of the major friction points with AI.

This means that by having a single source of truth you don’t need to move your data to reason over it.  For example, your Data Scientists and Subject Matter Experts (SMEs) can utilize a single source of data to label and train the datasets while business users and field operations can utilize that same data to perform AI based inspections and AI assisted reality modeling. 



First, most AI computer vision applications begins with managing and labeling the datasets.  This is a critical step but often overlooked by most people. Typically, hundreds and more often thousands of images are necessary for training.  While there are many stand-alone AI Labeling tools available,  these point solutions often require users to replicate that data which creates error prone data silos.  In addition this makes it challenging to share and collaborate especially on larger and complex datasets limiting the potential accuracy of the training data.

Second, the AI Labeling toolset needs to provide support for a wide range of AI labeling features such as Image Classification, Object Detection, and Object Segmentation.  In addition, the solution must allow subject matter experts to collaborate and label the datasets on a single platform

By integrating a robust data labeling toolset with an enterprise data management platform allows stakeholders to manage and label this data across the organization.  Thus eliminating one of the biggest challenges companies faces with managing their AI workflow.

This means that by having a single source of truth you don’t need to move your data to reason over it.  For example, your Data Scientists and Subject Matter Experts (SMEs) can utilize a single source of data to label and train the datasets while business users and field operations can utilize that same data to perform AI based inspections and AI assisted reality modeling.



Next, all that labeled data must be packaged and made easily accessible for the purpose of AI training and ultimately creating the final AI classifier model(s).  One of the major friction points today is the AI training and model creation process is very disjointed from the dataset, often requiring manually exporting the label data and manipulating the datasets in order to act and reason over it.

Providing an easy to access digital asset database and single source of AI label data can dramatically reduce the manual efforts required by Data Scientists and developers to build the AI Model.

In addition, AI computer vision is a rapidly evolving and dynamic field. The technology is evolving and there are many great AI companies and AI technologies that when properly harnessed together can solve very complex problems.  Thus, having an open architecture that allows leading AI partners as well as Enterprise’s own AI Data Scientists to build robust AI models that are all usable on a single open AI platform.  

This democratized approach allows Enterprises to address many applications, solve today problems, and help address future challenges. 



Next, the AI mode needs to be made easily accessible and usable.  In other words, it needs to be operationalized to be effective. Briefly, the resulting AI model by itself is just a library that is allows specific inference to occur on the image.  However, Enterprises must figure out how allow the business groups to easily consume and to make use of that AI in their operations.

Thus, deploying the AI model is essential to an effective AI process.  Enterprises are greatly benefited if Business Users and stakeholders can rapidly consume and put the AI models into use solving critical business needs. 

This means providing users an “AI Marketplace” that allows users to select the available AI model  that best fits their application and use case. In addition,  it is also important to reduce additional time and effort from the IT organization to create custom deploying scenarios every time a new AI model is created.



In order for these AI based analytics to provide true value to an enterprise, they must be deployed within a pipeline that connects a robust visual data management platform with a configurable workflow that execute purpose-built analytics against the right data set.  The results of the AI visual analysis must be tightly integrated into the operational workflows to drive business intelligence.

 When properly implemented AI allows efficient visualization of results.

  • Powerful dashboards and AI power results to quickly identify critical issues
  • AI Classifications are integrated into the digital inspection twin so that its searchable
  • Geo-visual insights are automatically associated with AI results to enable rapid asset location and identification
  • AI results integrated into business intelligence workflows to drive greater impacts

In addition, the visualization of the AI results should also be extendable to other workflows such as dispatch and ticketing.  For instance, events to schedule maintenance calls based on AI results and an accurate view of the current status of the physical asset is possible with a unified digital transformation platform.


In short, Enterprises are looking to seamlessly incorporate their AI libraries into their visualization workflow.   AI should not be a series of separate and disjointed processes.  AI reaches its full potential when it is put to work in concert and complement this visual data. 

Optelos can help you Operationalize and transform your AI program with our unified data visualization platform that seamlessly incorporates the AI Brain into the workflow.  We allow business users and decision makers to focus on the results, isolate issues, and take action instead of having to deal with all the plumbing require to use the AI.  In other words, we operationalize AI to allow you to analyze and visualize results in a single pane of glass.

Let the dataset change your mindset”  Hans Rosling

Contact Us To Learn More