Computer Vision Across Industries: From Lab to Production Line

RISE computer vision research spans a range of industrial applications, from biotechnology to materials manufacturing. Here are three strategic projects that illustrate how AI-based vision systems are transforming production processes.

Life Science Applications

In life science, neural networks have been shown to predict complex chemical properties from simple images, demonstrating the potential for real-time quality control in biomass production. By correlating standardized images of product vials with phosphorus NMR results using a simple PyTorch-based model, image analysis can serve as a fast decision-support approach that complements slower laboratory assays and illustrates a path toward scalable AI-assisted quality monitoring.

Construction Materials

In construction materials, our collaboration with Hasopor AB showed the potential of a low-cost system for measuring thickness and width in cellular glass production using 3D cameras and YOLOv11 segmentation. The now concluded project demonstrated that a 3D Intel RealSense camera can capture the cellular glass “cake” on the line and deliver measurements comparable to manual checks. While not on par with laser systems, the approach is accurate enough to be a viable and affordable alternative for production lines where lasers are deemed too costly.

Advanced Materials Characterisation

Meanwhile, AI-assisted analysis of small-angle X-ray scattering (SAXS) and X-ray computed tomography is enabling unprecedented depths of material characterisation, with rapid interpretation of complex structural data. Through Vinnova-funded projects with partners such as AstraZeneca and Tetra Pak, machine learning models are being developed to infer structural parameters directly from SAXS profiles. In collaboration with Billerud and Lund University, research is advancing automated instance segmentation models for cellulose tomography, which will substantially reduce manual effort, analysis time, and variability while enhancing overall robustness.

Impact

Together, these efforts show how computer vision has evolved from a laboratory tool into a practical, scalable industrial asset, improving efficiency, quality, and safety across multiple sectors.