Privacy-Aware Vision AI for Industry (PRIVAI)
This project evaluates privacy-preserving vision AI in manufacturing applications, including its technical performance, industrial viability and acceptance among workers.
However, these systems often capture sensitive worker and company data, raising privacy concerns that hinder widespread adoption of innovative computer vision solutions. IDLab Ghent and Flanders Make have developed a privacy-preserving AI technology that masks identifiable human attributes (e.g., tattoos, glasses) in a video, while ensuring that useful insights can still be extracted from the transformed video by downstream algorithms. While this approach has been shown to be effective for ergonomic risk assessment (for a video of the demonstrator, see below), further validation is needed in broader industrial applications and for work acceptance.
This project will evaluate the current technology in two real-world manufacturing use cases - operator activity tracking and forklift navigation – and conduct a user acceptance study to assess the conditions under which workers will accept being filmed at work. The results will support industry adoption of vision-based AI in manufacturing and identify key research challenges for follow-up research projects.
Downloads
Contacts
Abdellatif Bey-Temsamani (Flanders Make) - abdellatif.bey-temsamani@flandersmake.be
Researchers
Emilio Gamba (Flanders Make)- emilio.gamba@flandersmake.be
Research labs
Ugent - IDLab
Flanders Make