News

Classifying defects more reliably

January 8, 2020 - While AI-based technologies can now outperform humans in many categories of image processing, the "experienced" reliability of surface inspection systems in the steel industry often leaves much to be desired. In apparent contradiction to this, the very same systems are able to achieve high classification accuracy on controlled test data. A structured optimization of datasets and classifiers using deep learning technology drastically increases the practical performance of existing systems. For the best possible performance, surface defects must be classified multiple times throughout production.

Read more about the solutions that Smart Steel Technologies offers to address the issues surrounding surface inspection systems in the article published in STEEL + TECHNOLOGY in the article linked below.

Classifying-Defects-More-Reliably-Steel-and-Technology.pdf