An extra sense for gathering and analysing data

Share this post:

Ever considered inspecting a production line with an iPhone, a warehouse with a robot, or a bridge with a drone? Developers can now enhance applications with a ready-to-use module for image recognition: IBM Maximo Visual Inspection. This technology incorporates AI that analyses camera images to detect maintenance issues, steer autonomous vehicles or perform various other tasks. How does this module work in practice?

It is the first time that developers have access to a module of image-recognition software in which AI is an integral element. This makes it possible to equip solutions with an extra sense that not only signals, but also analyses. In this article, Ronald Teijken, Business Partner Manager Benelux for IBM MAS and Damiaan Zwietering, IBM Developer Advocate and Data Scientist, shine their light on the possibilities of this module, based on a number of practical examples.

Mature technology

‘The image-recognition technology behind Visual Inspection is not new. At IBM, we have been building up practical experience with it for years,’ says Ronald Teijken. ‘We have previously adapted it to enable self-driving vehicles to recognise obstacles and traffic signs. In fact, the technology has been available to developers as open-source software for even longer.’

In the meantime, the image recognition technology has matured to the extent that IBM has integrated it as part of the IBM Maximo Application Suite (MAS) and added AI as in integral part of it. On the platform, developers can include this combination of image recognition with AI as a separate module in their applications. Read more about the possibilities of IBM Maximo Application Suite (MAS) in the blog >

Artificial Intelligence

End-users who are working with a Visual Inspection application begin by tagging videos and photos. ‘For every image of a good situation, we put a little tick; we don’t add one to images of undesirable situations,’ explains Damiaan Zwietering. ‘Subsequently, the module can differentiate between ‘good’ and ‘wrong’. Users can adjust the settings under which they will receive notifications, and which parameters AI will take into account in its assessment, such as temperature. As well as this, they can indicate the values under which they wish to receive messages.’ In this way, Visual Inspector helps to indicate, more quickly and accurately, when a current camera image provides sufficient reason to take action, taking the repetitive inspection tasks out of the hands of users.

In practice

Image recognition in combination with AI can save time and improve product quality for various sectors. Visual Inspection is also used in diverse ways, from predictive maintenance with infrastructural objects to factory and warehouse inspections. To give an example, Toyota uses Maximo Visual Inspection to monitor production lines. How does it do that? ‘An iPhone takes photos on the production line throughout the day. Via an internal 5G network at the production location, these images are sent in real time to a data platform for analysis. Toyota is then in the position to quickly and accurately check thousands of factory parts every day,’ says Ronald Teijken.

From robot dog to autonomous ship

Another highly illustrative example is the manner in which Boston Dynamics uses Visual Inspection. The tech company has fitted its robot dog Spot with a camera that is linked to image-recognition technology. Just like its flesh-and-blood brothers, Spot is a loyal helper: the robot dog is used in the inspection of production facilities and warehouses. Spot is of greatest value in locations where there is a potential safety risk.

‘Even at sea, Visual Inspection is proving its power,’ says Damiaan Zwietering. ‘Thanks to this module, the ship, Mayflower, can sail entirely autonomously. The image recognition technology recognises buoys, other ships, bridges, driftwood, piers and more. Afterwards, AI combines the images with all sorts of other data to determine the right speed and direction. This includes the number of knots that other ships in the area are travelling, windspeed and water depth.’

Water purification

What is the water quality in a reservoir? ‘Water companies use Visual Inspection to be constantly updated on the water quality and can observe trends in the data,’ says Damiaan Zwietering. ‘For this, AI bases its assessment not just on the camera images it’s receiving, but also on the current and historic value of other parameters such as temperature.’

The Danish company Sund & Baelt uses the module to perform inspections, but on a somewhat different location: the bridge joining Denmark and Sweden, the third-longest one in the world. ‘Maintenance workers no longer have to come up to the top of high bridge pillars. Working at great heights is time-consuming and not without danger.’ This is why the bridge is inspected by camera drones that fly around it. Visual Inspection analyses the images and detects defects such as cracks or rust. Due to the benefits it offers for inspecting infrastructure objects, IBM has included Visual Inspection in the new solution, IBM Maximo for Civil Infrastructure.

The AI image recognition module from IBM is ‘pre-trained’. As a result, the output is of a very high standard from the start. This means that developers and data scientists waste less time on teaching and training the model until it is mature enough to be put to practical use. Plus, the module is self-learning, so the results become more accurate as more image data is analysed. However, 100 per cent accuracy is never possible without human intervention; human expertise is vital for fine-tuning the model to remove the last little errors from the output or perfectly adapting the algorithm to the industrial application. And that’s a nice challenge for developers and data scientists.

Would you like to know more about IBM Visual Insights?
Watch the video >

Start using Visual Inspection yourself:
Check here >