SynVis

Data is a significant bottleneck for the application of machine learning, with 80 % of the time spent on development being associated with the acquisition and annotation of large datasets, with costs in the millions of DKK. At SynVis we can quickly create digital environments and generate large datasets with realistic, annotated images. We make it faster and cheaper to implement machine learning.

Why?

Machine learning has immense potential to revolutionize automation, when combined with machine vision systems. The quality and size of datasets will be critical to the performance of AI, however, datasets are time-consuming and costly to make. If changes occur, like the introduction of a new product or use in a different environment, new datasets must be made from scratch, at great expense.

How?

We recreate a real-world environment digitally and use it to create synthetic images with annotations to augment or replace traditionally acquired images. We can easily provide variation by parameterizing conditions like lighting, settings, physical appearances and positions. With our experience as mechanical engineers working with automation, robotics and machine vision, we have insight into production and how machine vision systems fit in.