Detecting fabric defects with vision AI

Robovision developed an AI-based system that can spot fabric defects faster than manual inspection, while allowing Veranneman Technical Textiles, part of SIoen Industries, to keep the knowledge in-house.


Spotting defects during production — not after

Textiles are big in Belgium and Veranneman Technical Textiles, part of Sioen Industries, is a leader. Their wide range of high-performing tech fabrics serves companies around the world, from healthcare to the automotive and construction industries. With a variation in products this big, the variation in mistakes becomes equally challenging. They were looking for a solution to spot any hangup right when it’s happening and keep the extensive knowledge of product specialists on board, especially in a field scrambling for qualified personnel.

Figure 1: Veranneman Technical Textiles was looking for a solution that can spot any mistakes of various textile products hanging up right 

A smart quality control system

The solution we developed with our partner Viu More scans every millimetre of produced fabric in-line, in real-time. These images are then used for quality control, which is now faster and more accurate than ever before, covering every bit of fabric produced in real-time versus periodic quality control. Plus, our AI integrates with all the existing machinery, even weaving machines dating back to the 1960s. No need to replace a single gear.

What’s more: operators can easily train the vision software to perform its tricks on similar flaws, like colour degradation or thread thickness. By logging every single flaw or mistake as well as the machine’s parameters for a whole year, they’re building a database that soon will allow them to close the feedback loop and let the AI adjust presets as soon as it spots a certain kind of mistake.

Figure 2: smart cameras are used to scan every millimetre of produced fabric in-line, in real-time and feed images to the AI model to detect defects

Aligning operators and machines

Our AI makes line operators happy, too. They feed the machine with their knowledge and in return boost the end quality of every roll of fabric leaving the factory. In other words: the software allows operators to perform better. In an industry where skilled personnel is hard to find, we provide a solution that lets humans and machines work together, creating collaborative intelligence.

Figure 3: Operators can use their product knowledge to teach machines to see using the Robovision Platform

Reaching Desirable Outcomes


Which means faster action when a flaw is spotted, as well as decreased product loss.


Clients know exactly what they’re buying, courtesy of detailed reporting.


Gathering lots of data on defects allows for endless fine-tuning of production processes.


Operators can train a new model without needing large data sets.


Fewer flaws equal a superior fabric, giving Sioen Veranneman a head start on the competition.

Want to automate visual inspection?

Talk to an Expert