FEMM Hub Researcher presents at The Univeristy of York

FEMM Hub Research Assistant Ioan-Alexandru Herdea presented his paper 'Data Augmentation Framework for Improved Classification in Object Detectors' at the University of York on 20th May 2025. He had published the journal paper in IEEE Access during his PhD in “in-process vision-based quality inspection for manufacturing of components with inner features” and was invited to talk about it by Dr. John Oyekan, a Senior Lecturer in the Department of Computer Science and one of the authors in the article, facilitated the talk to be delivered in person and online to fellow staff members and Master’s students. 

The talk started with a short introduction on deep learning for quality inspection for the manufacturing of electrical machines. It was then followed by an explanation on why choosing an appropriate pixel-level augmentation technique is considered a research gap when object detectors have insufficient training images. The discussion went on about the methodology used and the observations after comparing various experimental results. The presentation culminated with the description of a novel pixel augmentation framework with the emphasis that it serves as a methodology rather than a suits-all-cases tool. With that, questions followed from several people in the room covering a wide range of perspectives, from technical information to future work plans and possible outcomes when extrapolating the framework to their specific use cases. At that point, the discussion felt more like a conversation than a Q&A. Alex took the questioning part as an opportunity to practice for his PhD viva, which is due soon, and made a note of each question to cover it in his PhD thesis. At the end of the talk, he  asked the audience for feedback on his  presentation skills, and the answers were above his expectations.