Danijel Skočaj
Associate Professor, University of Ljubljana
Bio:
Danijel Skočaj is a Full Professor at the University of Ljubljana, Faculty of Computer and Information Science, and the Head of the Visual Cognitive Systems Laboratory. His primary research interests lie in the fields of computer vision, machine learning, and cognitive robotics. Within the scope of basic and applied research, he has been developing new advanced methods in deep learning and computer vision to tackle complex problems that require the processing of visual information. He is also interested in the ethical considerations of artificial intelligence, machine learning, and robotics, and the impact of these technologies’ development on society. Prof. Skočaj lectures on topics within the realms of computer vision, cognitive robotics, and deep learning. He has led or contributed to numerous projects in these research areas, including EU projects, national research projects, and industry-funded applied projects. His involvement in applied research projects has facilitated the transfer of research discoveries into practical applications. Prof. Skočaj has served as President of the IEEE Slovenia Computer Society and the Slovenian Pattern Recognition Society.
Lecture:
Data-driven deep-learning-based machine vision
Abstract:
Over the past decade, the field of computer vision has witnessed monumental strides, primarily driven by the renaissance of deep learning, and significantly fuelled by the increasing accessibility of data. In recent years, this data-driven, learning-based approach has started permeating into the traditionally conservative engineering discipline of machine vision. This paradigm has shown substantial promise as a potential alternative to the conventional methodology which relies heavily on hand-engineered, problem-specific solutions. The learning-based methodology advocates for a more general, efficient, flexible, and cost-effective development, deployment, and maintenance of machine vision systems. In this talk, we intend to delve into this emergent development paradigm. We will elucidate several data-driven approaches to surface defect detection and some other machine vision tasks. We will consider various learning paradigms, ranging from unsupervised techniques to fully supervised methods. We will discuss the advantages of these approaches and the challenges they face and address the role and opportunities of learning-based approaches for efficient visual inspection as well as for solving other tasks that rely on the processing of visual information in the framework of the Industry 4.0 paradigm.
