New deep learning

The new Deep Learning add-on for Adaptive Vision Studio is an intuitive and reliable solution for machine vision. It was developed for problems that, until now, were hard to solve or couldn’t be solved at all with conventional image processing methods. Users are able to create machine vision solutions without any programming experience. The software is trained using good and bad samples. The objects are then automatically recognized as such.
The requirement for typical applications are approx. 20 to 50 images for learning purposes, a powerful up-to-date GPU, and a few minutes of training time. Users are presented with reliable inspection results for their application. The add-on is offered in both a supervised and unsupervised mode for better results. In the supervised mode, users can mark defective pixels themselves. The add-on then learns the main properties of these defects. This can lead to better results, particularly for inspection tasks in the textile, photovoltaics, food and other sectors.
Further information is available under “Software”.