Model-Assisted Labeling Deployment – Enablers for AI and IOT Paving Modern Iron- and Steelmaking

With the advance of Artificial Intelligence (AI) and internet of things (iOT) devices product quality increase, throughput maximization and process stabilization in metals industry are achieved. Hardware capabilities and mathematical knowledge provide new digitalization methodologies. One of the most prominent areas is computer vision. Primetals Technologies offers Data Analytics for Computer Vision as a service. It involves dealing with robustness of algorithms for Computer Vision in the harsh and changing environmental conditions of steel plants, considering disturbing effects such as changing weather conditions. On the other side, separate model development and labelling workflows are error-prone and time-consuming. Within the scope of this paper, we show our workflow of applying robust Machine Learning methods, dealing with harsh outdoor conditions such as rain and snow and combining both learning, labelling and deployment. We present the advantages through algorithm performance and development time. In addition to that, we show the importance of semi supervised learning for Computer Vision by well-established use cases in scrap composition and crack detection.

Adnan Husakovic | Primetals Technologies Austria GmbH
 

Ali Abbas | Primetals Technologies Austria GmbH
 

Andreas Melcher | Primetals Technologies
 

Anton Tushev | Primetals Technologies Germany GmbH
 

Thazhath Johnce Swathish | Primetals Technologies Austria GmbH